Objectives: The aim of this study was to compare the performance of the Glasgow-Blatchford and the AIMS65 scoring systems as early risk assessment tools for accurately identifying patients with upper gastrointestinal (GI) bleeding who are at a low risk of requiring clinical interventions, including emergency endoscopy. The secondary objective was to compare their performance regarding relevant clinical outcomes.
Methods: Data were collected prospectively over a 2-year period in the emergency department of a university hospital. Adult patients with upper GI bleeding from either variceal or nonvariceal sources were included. Composite clinical outcomes consisted of a need for surgical or endoscopic intervention, rebleeding, intensive care unit admission, or in-hospital mortality. Patients who required blood transfusions or suffered composite clinical outcomes were considered high-risk patients. Glasgow-Blatchford score (GBS) and AIMS65 score were calculated for each patient. The sensitivity and specificity of the scoring systems were calculated. The areas under the receiver-operating characteristic curve (AUC) of the scores were compared.
Results: There were 254 patients in the study, of whom 163 (64.2%) were men. The median age was 61 years (interquartile range = 45 to 72 years). Among the patients, 211 (83.1%) underwent endoscopy, of whom 49 (19.3%) required endoscopic intervention to achieve hemostasis. Five (2%) patients required surgical intervention. Rebleeding was observed in 33 (13%) patients. A total of 143 (56.3%) patients received blood transfusions. A total of 152 (59.8%) were defined as high risk. Eighty-one (31.9%) experienced at least one component of the composite clinical outcomes, 18 (7.1%) of whom suffered in-hospital mortality. A GBS of 0 was observed in 16 patients (6.3%) in the study group. Two of these were high-risk patients. A total of 101 (39.8%) patients had AIMS65 scores of 0. Thirty-four of these were high-risk patients. A GBS of 0 had higher sensitivity than an AIMS65 score of 0 (98.68% vs. 77.6%). The negative predictive values of the GBS and AIMS65 of 0 were 87.5 and 66.3%, respectively. The GBS and AIMS65 were similar with regard to the composite outcome prediction, with AUCs of 0.795 (95% confidence interval [CI] = 0.74 to 0.843) and 0.746 (95% CI = 0.688 to 0.798), respectively (p = 0.137). The scores were also similar with respect to predicting in-hospital mortality (AUCs of 0.85 vs. 0.81; p = 0.342). The GBS was superior to the AIMS65 in identifying high-risk patients, with AUCs of 0.896 (95% CI = 0.85 to 0.93) and 0.771 (95% CI = 0.714 to 0.821; p < 0.001), respectively. The GBS was also more accurate than the AIM65 in predicting the need for blood transfusions (AUCs of 0.904 vs. 0.796; p < 0.001) and interventions (AUCs of 0.727 vs. 0.647; p = 0.05).
Conclusions: These results suggest that the GBS has superior sensitivity relative to the AIMS65 in identifying patients who were not likely to require interventions, including emergency endoscopy. Additional work to determine the use in real-time decision making may be warranted and helpful in providing guidance to clinicians.
© 2014 by the Society for Academic Emergency Medicine.