Read the full EBM Focus and earn CME credit.
Reference: BMJ 2017 Jan 4;356:i6432 (level 2 [mid-level] evidence)
- Five risk assessment tools (Rockall, Progetto Nazionale Emorragia Digestiva, pre-endoscopic Rockall, Glasgow Blatchford, and AIMS65) were evaluated for their ability to predict a composite outcome of hospital-based intervention or death in patients with upper gastrointestinal bleeding.
- The Glasgow Blatchford risk tool had the best predictive performance, and a cutoff score of ≤ 1 was determined to be the optimal threshold for identifying patients at low risk for the composite outcome.
- In 564 patients identified as low risk using the Glasgow Blatchford tool, 3.4% required a hospital-based intervention or died.
An American College of Gastroenterology conditional recommendation suggests that patients with upper gastrointestinal (UGI) bleeding who have a Glasgow Blatchford risk score = 0 may be considered for discharge from the hospital without inpatient endoscopy (Am J Gastroenterol 2012 Mar;107(3):345). However, there are other risk prediction scoring systems, and so to determine the best method for identifying low-risk patients, a prognostic cohort study evaluated the performance of five risk assessment tools. The goal was to be able to select patients with UGI bleeding who may be suitable for outpatient management. Calculation of two of the risk tools, the full Rockall and Progetto Nazionale Emorragia Digestiva (PNED), requires endoscopy; the other three risk tools, initial (pre-endoscopic) Rockall, Glasgow Blatchford, and AIMS65 bedside risk, do not. This study assessed all five risk tools in 3,012 consecutive patients with UGI bleeding presenting to six participating hospitals located in the USA, Scotland, England, Denmark, Singapore, and New Zealand and followed them for 30 days. The composite outcome of hospital-based intervention (red blood cell transfusion, endoscopic treatment, interventional radiology, or surgery) or death occurred in 45% of patients. Performance of all five risk tools for the prediction of the composite outcome was calculated in 1,704 patients (57%) for whom there was complete data for both the risk tools and outcomes. Optimal cutoff values were selected to identify low risk patients who may be suitable for outpatient management.
Among the five risk assessment tools, the Glasgow Blatchford tool had the best performance to discriminate for the composite outcome. In analysis of the separate outcomes, the PNED tool had the best performance to predict rebleeding and both the PNED and AIMS65 had the highest performance to predict mortality. In an evaluation of the three risk tools that do not require endoscopy, the risk of a hospital-based intervention or dying in patients with low risk scores was 3.4% for patients with Glasgow Blatchford score ≤ 1, 14% for patients with initial Rockall score = 0, and 25% for patients with AIMS65 score = 0. In patients identified as low risk with each tool, all-cause mortality was 0.4% with the Glasgow Blatchford, 0.2% with the initial Rockall, and 0.7% with the AIMS65.
In this study, the Glasgow Blatchford tool performed the best to discriminate high- vs. low-risk patients with UGI bleeding. However, this tool did not perform the best to predict each of the separate outcomes, suggesting that use of more than one tool may be necessary depending on the outcome of focus. The performance of the Glasgow Blatchford tool using the cutoff value of ≤ 1 point to rule out patients at high risk of hospital-based intervention or death shows that it can be used to help identify patients who may be suitable for emergency room discharge with outpatient follow-up rather than admission. Although the Glasgow Blatchford tool has been evaluated in other studies, the selection of this cutoff value (≤ 1) and determination of the corresponding risk (3.4%) were conducted in the same cohort and still require validation in a separate cohort of patients. Pending validation, these results suggest that the cutoff using the Glasgow Blatchford risk tool could be raised from 0 to ≤ 1 for identifying patients with an UGI bleed who are at low risk for serious complications. For identifying such patients, the Glasgow Blatchford risk tool outperforms the other risk assessment tools.
For more information, see the Acute upper nonvariceal gastrointestinal bleeding topic in DynaMed Plus. DynaMed users click here.