Comparison of risk stratification scores for patients presenting with symptoms of upper gastrointestinal bleeding in the emergency department

Objective: To compare four scoring systems to predict outcomes in patients with symptoms of upper gastrointestinal bleeding presenting to the emergency department. Method: A single centered prospective cohort study. All adult patients presenting to the emergency department of the studying centre with haematemsis or tarry stool or coffee ground vomiting or coffee ground aspirate from nasogastric tube were included from February 2012 to April 2012. The outcome variables include mortality, length of stay in hospital, blood product transfusion and interventions for bleeding control. The AIMS65 score, pre-endoscopic Rockall score, Glasgow Blatchford Score (GBS) and the modified Glasgow Blatchford Score (mGBS) were evaluated. Diagnostic characteristics were presented and areas under the receiver-operating-characteristic (AUROC) curve were compared. Results: A total of 129 patients were included in the study. 81 of them (62.8%) had upper endoscopy performed. The mortality rate was 3.1%. Initial haemoglobin level of <10 was an important factor in risk stratification. Validation of the 4 scoring systems showed GBS had highest sensitivities (98.3-100%) and negative predictive values (90-100%) for all outcome variables but could not achieve a good specificity and positive predictive values against the outcomes. Both GBS and modified GBS outperformed the other two scoring systems in the AUROC curves in predicting composite high-risk outcome, length of stay in hospital and blood transfusion. Conclusion: GBS appeared the best scoring system in the emergency department for screening purpose and to stratify those high risk patients for admission and low risk patients for out-patient management.