Risk Assessment Scores for Upper GI Bleeding
The Glasgow Blatchford Score (GBS) is the preferred prognostic tool for upper GI bleeding, with superior sensitivity (99%) in identifying high-risk patients and excellent ability to identify low-risk patients suitable for outpatient management. 1
Primary Risk Stratification Scores
Glasgow Blatchford Score (GBS) - First-Line Tool
Use the GBS as your initial risk assessment tool for all patients with upper GI bleeding. 1
The GBS includes only pre-endoscopy clinical and laboratory data: 2, 1
- Hemoglobin level
- Blood urea nitrogen level
- Systolic blood pressure
- Heart rate
- Presence of syncope
- Presence of melena
- Evidence of hepatic disease
- Evidence of cardiac failure
A GBS ≤1 identifies patients at very low risk who can be safely managed as outpatients without early endoscopy (sensitivity 98.6%, specificity 34.6%). 1, 3 This threshold has been validated in large international studies and accurately predicts survival without need for intervention. 3
A GBS ≥7 is the optimal threshold for predicting need for endoscopic treatment (sensitivity 80%, specificity 57%). 3
The GBS outperforms all other pre-endoscopy scores with an AUROC of 0.86 for predicting intervention or death, compared to admission Rockall (0.66) and AIMS65 (0.68). 3
Rockall Score - Mortality Prediction
The Rockall score exists in two forms: 2, 1
Pre-endoscopy Rockall: Uses only clinical data (age, shock, comorbidities) before endoscopy. This has limited utility with AUROC of only 0.66 for predicting intervention. 3
Complete (post-endoscopy) Rockall: Adds endoscopic findings (diagnosis, stigmata of recent hemorrhage). 2
Use the complete Rockall score specifically for mortality prediction (AUROC 0.72), where it performs better than GBS (AUROC 0.64) but remains inferior to AIMS65. 3 The Rockall score has better discriminative ability for mortality than for rebleeding. 2, 1
A complete Rockall score ≥5 predicts mortality with sensitivity 65.8-78.6% and specificity 65.0-65.3%. 3
The Rockall score may misclassify 4-7% of high-risk patients as low risk, limiting its use for safe discharge decisions. 1
AIMS65 Score - High-Risk Mortality Identification
Do not use AIMS65 for identifying low-risk patients suitable for discharge. 1 The American College of Physicians specifically recommends against this application. 1
AIMS65 is designed primarily to identify patients at high risk for death, with AUROC of 0.77 for mortality prediction (equal to PNED score and superior to Rockall). 3
An AIMS65 score ≥2 predicts mortality with sensitivity 65.8-78.6%. 3
However, AIMS65 has lower sensitivity (78-82%) for identifying high-risk patients compared to GBS, making it less useful for initial triage. 1
Clinical Application Algorithm
Step 1: Calculate GBS on presentation using readily available clinical and laboratory data. 1, 3
Step 2: If GBS ≤1, consider outpatient management without urgent endoscopy, accounting for access to care, urban vs rural setting, and patient preferences. 1
Step 3: If GBS ≥7, anticipate need for endoscopic intervention and prepare accordingly. 3
Step 4: After endoscopy, calculate complete Rockall or AIMS65 if mortality risk stratification is needed for prognostication or family discussions. 3
Important Caveats
No score accurately predicts rebleeding. 3, 4 All available scores have AUROC <0.80 for this outcome, limiting their clinical utility for predicting which patients will rebleed. 3
No score accurately predicts length of hospital stay. 3
The GBS has been validated internationally across multiple continents with consistent performance. 3 However, one US-based study found suboptimal specificity at all thresholds, though GBS still outperformed pre-endoscopy Rockall. 5
When using GBS for discharge decisions, consider patient-specific factors including distance from hospital, access to emergency services, availability of urgent endoscopy, and patient ability to return if symptoms worsen. 1