Written by Ketan Patel and Kyle Rankin
Spoon Feed
For patients in the ED with upper GI bleeding (UGIB), the AIMS65 score best predicted mortality, but none of the scores were accurate enough to be clinically useful.
UGIB crystal ball…
UGIB is common, dangerous, and hard to triage quickly, so clinicians need accurate tools to spot who’s likely to die in hospital. We use scores like Glasgow-Blatchford, Pre-Endoscopic Rockall, and AIMS65. Early warning scores like NEWS-2 have not been directly compared in this setting.
In a prospective observational study of 178 adults with confirmed UGIB presenting to the ED, researchers compared the prognostic performance of NEWS-2 and its lactate-enhanced version (NEWS2-L) with conventional scores to predict in-hospital mortality, which occurred in 20.2% of the population. AIMS65 had the highest sensitivity and lowest negative likelihood ratio, while NEWS-2 had highest specificity and strongest positive likelihood ratio. NEWS2-L offered only minimal improvement. See table. In multivariate analysis, only AIMS65 and NEWS-2 remained independent predictors of death.
Limitations include assessment as a single-center and a modest sample size, limiting generalizability.
How does this change my practice?
In practice, calculating both AIMS65 and NEWS-2 for ED patients with UGIB may sharpen early mortality risk stratification—helping decide who needs urgent endoscopy, ICU care, or closer monitoring. We’ve already had components of NEWS-2 embedded in early warning systems for sepsis in our EHR, and I anticipate wider use of tools such as these to help guide early identification and thus intervention on subsets of patients.
Editor’s note: None of these scores are very accurate. Don’t lean too heavily on them for decision making. ~Clay Smith
Source
Mortality prediction among ED patients with upper gastrointestinal bleeding: Comparison of NEWS-2 and conventional risk scores. Am J Emerg Med. 2025 Dec 21;101:14-21. doi: 10.1016/j.ajem.2025.12.020. Epub ahead of print. PMID: 41448104.
View JournalFeed Critical Appraisal
Critical Appraisal
Study Identification
Background
Study Question
Study Design & Conduct
Prospective / Retrospective: Prospective
Multicenter: No
Unit of Allocation: Not applicable
Unit of Analysis: Patients
Randomization Method: Not applicable
Allocation Concealment: Not applicable
Blinding: Not applicable
Follow-up Duration: In-hospital
Population
- Adult patients (≥18 years)
- Clinically confirmed upper gastrointestinal bleeding
- Age under 18 years
- Pregnancy
- Diagnosis of lower gastrointestinal bleeding
- Incomplete or missing medical records
- Refusal to participate
Number Enrolled: 178
Number Analyzed: 178
Key Baseline Characteristics
Sex: 61.8% male
Disease Severity: Not reported
Care Setting Distribution: Emergency department
Additional Baseline Characteristics
- Higher respiratory rates and lower oxygen saturation in non-survivors
- Comorbidities like CHF and malignancy more frequent in non-survivors
Exposures / Interventions
Description: Calculation of NEWS-2 and NEWS2-L scores
Definition / Dose: Not applicable
Timing: At presentation
Classification Method: Protocolized
Protocolized / Discretionary: Protocolized
Description: Conventional scoring systems (AIMS65, GBS, PERS)
Definition: Standard definitions as per original scoring systems
Outcomes & Results
Primary Outcomes
Definition: Accuracy of scoring systems in predicting mortality
Time Point: In-hospital
Measurement Method: ROC curve analysis
Results: AIMS65 sensitivity 83.3%, NEWS-2 specificity 82.0%, NEWS2-L AUC 0.772
Secondary Outcomes
Definition: Not applicable
Time Point: Not applicable
Measurement Method: Not applicable
Results: Not applicable
Risk of Bias
Risk of Bias - ROBINS-I
- Confounding (Some concerns): Potential confounders like comorbidities were considered, but single-center design limits generalizability.
- Selection of Participants (Low): Clear inclusion and exclusion criteria were applied.
- Classification of Interventions (Low): Interventions (scoring systems) were clearly defined and applied consistently.
- Deviations from Intended Interventions (Low): No deviations reported; scoring was protocolized.
- Missing Data (Some concerns): Exclusion of patients with missing data could introduce bias.
- Measurement of Outcomes (Low): Mortality was objectively measured.
- Selection of the Reported Result (Low): All relevant outcomes were reported.
Transparency
COI Statement Present: TRUE
