About the Glasgow-Blatchford Score (GBS)
The Glasgow-Blatchford Score predicts whether a patient with upper gastrointestinal bleeding will need hospital intervention — transfusion, endoscopic therapy or surgery — using only admission observations and bloods, with no endoscopy required. It runs from 0 to 23. A score of 0 identifies patients at very low risk who can reasonably be managed as outpatients, which is what the score is mainly used for; 1–3 is low risk, and 4 or above marks patients who need urgent inpatient endoscopy. In the derivation cohort it achieved an ROC area of 0.92 for need for treatment.
Formula
GBS = urea points + haemoglobin points (sex-specific) + systolic BP points
+ pulse ≥ 100 (1) + melaena (1) + syncope (2)
+ hepatic disease (2) + cardiac failure (2)- Urea points
- mmol/L: ≥ 25 → 6; 10–24.9 → 4; 8–9.9 → 3; 6.5–7.9 → 2; < 6.5 → 0.
- Haemoglobin points (men)
- < 10 g/dL → 6; 10–11.9 → 3; 12–12.9 → 1; ≥ 13 → 0.
- Haemoglobin points (women)
- < 10 g/dL → 6; 10–11.9 → 1; ≥ 12 → 0.
- Systolic BP points
- < 90 mmHg → 3; 90–99 → 2; 100–109 → 1; ≥ 110 → 0.
- Clinical features
- Pulse ≥ 100 adds 1; melaena adds 1; syncope adds 2; hepatic disease adds 2; cardiac failure adds 2.
- Everything needed is available before endoscopy, which is the whole point — this is a pre-endoscopy score.
- Urea is scored in mmol/L. If your laboratory reports BUN in mg/dL, multiply by 0.357 to get urea in mmol/L, or use the unit toggle on this page.
- The maximum is 23. There is no upper cap applied beyond what the components allow.
Interpreting the result
The critical threshold is zero. A GBS of 0 identifies a group in whom the risk of needing intervention is low enough that outpatient management with early endoscopy is endorsed by guidelines, and that is the decision this score exists to support. Above 0, the score grades urgency rather than giving a discharge option: 1–3 is low risk but not the zero-risk group, and 4 or more indicates a substantially higher likelihood of needing transfusion or endoscopic therapy. Because the outcome modelled is need for treatment, do not read a low score as a low risk of death — for that, calculate Rockall after endoscopy.
| Score | Band | What it means | Action |
|---|---|---|---|
| 0 | Very low risk | Very low likelihood of needing transfusion, endoscopic therapy or surgery | Outpatient management with early outpatient endoscopy is supported by guidelines |
| 1–3 | Low risk | Low likelihood of needing intervention, but not the zero-risk group | Admission decision remains clinical; endoscopy within 24 hours |
| ≥ 4 | High risk | Substantially increased likelihood of needing transfusion, endoscopic intervention or surgery | Admit, resuscitate, group and save, endoscopy within 24 hours; consider Rockall afterwards for mortality risk |
Scroll the table sideways for every column.
What the Glasgow-Blatchford needs (9 inputs)
- Blood urea
- Scored in mmol/L: 6 points at 25 or above, 4 at 10–24.9, 3 at 8–9.9, 2 at 6.5–7.9, otherwise 0. Urea rises in upper GI bleeding from absorbed blood protein as well as from hypovolaemia, which is why it carries the heaviest weighting in the score. Enter BUN in mg/dL or urea in mmol/L using the unit toggle.
- Haemoglobin
- Thresholds differ by sex. In men: 6 points below 10 g/dL, 3 at 10–11.9, 1 at 12–12.9. In women: 6 points below 10, 1 at 10–11.9. Remember that haemoglobin lags acute bleeding — a normal value early does not exclude significant blood loss.
- Systolic blood pressure (mmHg)
- 3 points below 90, 2 points at 90–99, 1 point at 100–109, otherwise 0.
- Pulse (bpm)
- 1 point at 100 or above. Beta blockers blunt this response, so a normal pulse in a treated patient is less reassuring than it looks.
- Sex
- Used only to select the haemoglobin thresholds, which are more forgiving of a lower haemoglobin in women.
- Melaena on presentation
- 1 point.
- Syncope
- 2 points. A marker of the haemodynamic significance of the bleed.
- Known hepatic disease
- 2 points. Raises the possibility of variceal bleeding, which behaves differently and carries higher mortality.
- Cardiac failure
- 2 points. Both a marker of comorbidity and a constraint on how aggressively the patient can be resuscitated.
Units. Urea is scored in mmol/L, which is how most of the world reports it. If your laboratory reports blood urea nitrogen (BUN) in mg/dL — common in the United States — switch the unit toggle rather than entering the mg/dL figure directly: urea in mmol/L equals BUN in mg/dL multiplied by 0.357, and entering mg/dL as though it were mmol/L will under-score the patient roughly threefold. Haemoglobin can be entered in g/dL or g/L using its own toggle.
What it returns
- Glasgow-Blatchford Score (0–23)
- The sum of all weighted components. The maximum of 23 is reached only with extreme values across every domain.
- Risk of needing intervention
- The band, and what it implies for admission and endoscopy timing. Note the outcome predicted is need for treatment, not death.
How it is calculated
Blatchford and colleagues built the score from a Scottish audit of patients admitted with upper gastrointestinal haemorrhage, using multivariate analysis to find which admission variables predicted a composite need for treatment — blood transfusion, endoscopic intervention, or surgery. Each retained variable was given an integer weight roughly proportional to its regression coefficient, which is why urea carries up to 6 points and melaena only 1. The design constraint was deliberate: only variables available at first assessment were allowed, so the score could be used to make the admission decision rather than to review it afterwards.
Facts & figures
| Variable | Threshold | Points |
|---|---|---|
| Blood urea (mmol/L) | ≥ 25 / 10–24.9 / 8–9.9 / 6.5–7.9 | 6 / 4 / 3 / 2 |
| Haemoglobin, men (g/dL) | < 10 / 10–11.9 / 12–12.9 | 6 / 3 / 1 |
| Haemoglobin, women (g/dL) | < 10 / 10–11.9 | 6 / 1 |
| Systolic BP (mmHg) | < 90 / 90–99 / 100–109 | 3 / 2 / 1 |
| Pulse ≥ 100 bpm | Present | 1 |
| Melaena | Present | 1 |
| Syncope | Present | 2 |
| Hepatic disease | Present | 2 |
| Cardiac failure | Present | 2 |
Scroll the table sideways for every column.
Urea carries the greatest weight because upper GI bleeding raises it through absorbed blood protein as well as hypovolaemia.
Evidence
Derivation — Scottish audit of upper GI haemorrhage
2000 · n = 1,7481,748 patients admitted with upper gastrointestinal haemorrhage, with a separate validation set of 197 patients. The modelled outcome was need for treatment, defined as blood transfusion, endoscopic intervention or surgery.
ROC curve area 0.92 (95% CI 0.88–0.95) for need for treatment, with good calibration (p = 0.84 for patients needing treatment).
Adoption for outpatient management
2019Subsequent validation established that patients scoring 0 can be managed as outpatients with early endoscopy, and this pathway is endorsed in international consensus guidance on non-variceal upper gastrointestinal bleeding.
Consistently high sensitivity for identifying patients who will need intervention, at the cost of low specificity — the score is designed to be safe rather than efficient.
Comparison with Rockall
2019Head-to-head comparisons in upper GI bleeding cohorts have separated the two scores by the outcome each predicts best.
GBS was superior for transfusion need (AUC 0.810 versus 0.675 for Rockall in one cohort) and for detecting rebleeding, ICU admission and endoscopic intervention, while Rockall was superior for mortality (AUC 0.806 versus 0.750, and 0.842 versus 0.622 in another cohort).
How it compares
Glasgow-Blatchford vs Rockall Score
Use Glasgow-Blatchford before endoscopy to decide on admission and intervention, and Rockall after endoscopy to estimate mortality and rebleeding — they answer different questions and the evidence separates them cleanly.
GBS needs no endoscopic findings and predicts need for treatment; Rockall requires the endoscopic diagnosis and stigmata of recent haemorrhage and predicts death and rebleeding. In direct comparisons GBS outperformed Rockall for transfusion need (AUC 0.810 versus 0.675) and for identifying patients needing intervention, while Rockall outperformed GBS for mortality (0.806 versus 0.750 in one cohort, 0.842 versus 0.622 in another). The practical split is therefore sequential rather than competitive: GBS for the triage decision, Rockall for prognosis once the scope has been done.
Glasgow-Blatchford vs AIMS65
AIMS65 is quicker to calculate and better at predicting mortality; GBS is better at predicting the need for intervention, which is usually the more actionable question at presentation.
AIMS65 uses five equally weighted variables — albumin below 3.0 g/dL, INR above 1.5, altered mental status, systolic BP 90 mmHg or below, and age over 65 — so it is faster at the bedside. Studies generally find AIMS65 at least as good as GBS for in-hospital mortality and worse for transfusion and intervention need. Neither identifies a discharge-safe group as reliably as a GBS of 0.
Pearls & pitfalls
- The discharge threshold is 0, not a low number. A GBS of 1 is not the same decision as a GBS of 0, and the outpatient pathway was validated at zero.
- Urea is scored in mmol/L, not BUN in mg/dL. Confusing the two under-scores the patient by roughly a factor of three — use the unit toggle rather than converting mentally.
- Haemoglobin lags acute haemorrhage. A patient bleeding briskly for an hour can have a normal haemoglobin and score low on that component.
- Haemoglobin thresholds are sex-specific: 11 g/dL scores 3 points in a man and 1 in a woman. This is intended, not an inconsistency.
- Beta blockers blunt the tachycardia the score relies on, so a normal pulse in a treated patient is weaker reassurance.
- It predicts need for treatment, not death. Do not use it to reassure about mortality — Rockall discriminates that better.
- Known hepatic disease scores 2 points, but the score does not distinguish variceal from non-variceal bleeding, and variceal bleeding carries higher mortality than the number suggests.
Critical actions
- Calculate it before endoscopy — using it retrospectively wastes the only thing it is designed for.
- Discharge only at 0, and only with an arranged early outpatient endoscopy and clear return instructions.
- Resuscitate according to the clinical picture, not the score; a patient with a low score who looks unwell is unwell.
- Group and save early in anyone scoring 4 or above, since transfusion is the most likely intervention.
- Consider variceal bleeding separately in anyone with hepatic disease, and start vasoactive therapy and antibiotics if suspected rather than waiting for endoscopy.
- Calculate Rockall after endoscopy when mortality and rebleeding risk are the questions.
Why this score exists
The design decision that defines this score is the choice of outcome. Most risk scores in gastrointestinal bleeding predicted death, which is uncommon and therefore hard to act on at first presentation. Blatchford and colleagues instead modelled need for treatment — transfusion, endoscopy or surgery — because that is what determines whether a patient must be in hospital. The corollary is that the score is intentionally over-sensitive: it is built to avoid missing anyone who needs intervention, accepting that many patients scoring above zero will not need any. Reading a high GBS as a prediction of death, or a low one as reassurance about mortality, inverts what the authors set out to measure.
About the creator
First author, 2000 derivation study
Derived a risk score deliberately requiring no endoscopic data, so that it could be applied at first presentation.
Limitations
- Predicts need for treatment, not mortality, so it should not be used for prognostic conversations.
- Deliberately over-sensitive: most patients scoring above 0 will not need intervention, so specificity is low and it will not reduce admissions beyond the zero-score group.
- Does not distinguish variceal from non-variceal bleeding, despite the very different management and mortality.
- Haemoglobin lags acute blood loss, so early presentation can produce a falsely reassuring component score.
- Derived in a Scottish audit population in the late 1990s, before current proton pump inhibitor use, restrictive transfusion thresholds and endoscopic techniques.
- Physiological components are blunted by common medication — beta blockers on pulse, antihypertensives on blood pressure.
- Urea is affected by renal impairment and dehydration independently of bleeding, and carries the heaviest weight in the score.
If you are the patient
If you have come to hospital vomiting blood or passing black stools, this score helps the team decide how urgently you need treatment and whether you need to stay in hospital at all. It is worked out from your blood pressure, pulse, a blood count and a kidney blood test, plus a few questions about what happened — whether you fainted, and whether you have liver or heart problems. A score of zero means the risk of needing a blood transfusion or a procedure is very low, and it is usually safe to go home with a camera test (endoscopy) arranged soon afterwards as an outpatient. A higher score means you should stay, be monitored, and have the camera test within about a day. One thing worth knowing: this score is about whether you need treatment, not about how likely you are to die — those are different questions, and your team uses a different score for the second one, usually after the camera test.
Frequently asked questions
What is the Glasgow-Blatchford score?#
The Glasgow-Blatchford Score predicts whether a patient with upper gastrointestinal bleeding will need hospital intervention — transfusion, endoscopic therapy or surgery. It uses blood urea, haemoglobin, systolic blood pressure, pulse, sex and four clinical features, all available before endoscopy, and runs from 0 to 23.
What Glasgow-Blatchford score is safe for discharge?#
Zero. A GBS of 0 identifies patients at very low risk of needing intervention, and outpatient management with early outpatient endoscopy is supported by international guidance for that group. A score of 1 does not carry the same evidence, so the threshold is genuinely zero rather than 'low'.
What is the maximum Glasgow-Blatchford score?#
Twenty-three, reached only with extreme values across every component: urea of 25 mmol/L or more (6), haemoglobin below 10 g/dL (6), systolic BP below 90 (3), tachycardia (1), melaena (1), syncope (2), hepatic disease (2) and cardiac failure (2).
Is the Glasgow-Blatchford score better than Rockall?#
For different things. GBS is better for predicting the need for transfusion and intervention — AUC 0.810 versus 0.675 for transfusion in one comparison — and needs no endoscopy. Rockall is better for mortality, with AUC 0.806 versus 0.750 in one cohort and 0.842 versus 0.622 in another. Use GBS at presentation and Rockall after endoscopy.
How accurate is the Glasgow-Blatchford score?#
In the derivation cohort of 1,748 patients it achieved an ROC area of 0.92 (95% CI 0.88–0.95) for need for treatment, with good calibration. Its accuracy is asymmetric by design: very high sensitivity for identifying patients who need intervention, and low specificity, so most patients above zero will not need anything.
Does the Glasgow-Blatchford score need endoscopy?#
No, and that is the point. Every component is available at first assessment, so the score can inform the admission and endoscopy-timing decision rather than being calculated after the fact. Rockall, by contrast, needs the endoscopic diagnosis.
Why is urea weighted so heavily in the Glasgow-Blatchford score?#
Because in upper gastrointestinal bleeding urea rises for two reasons at once: hypovolaemia reduces renal perfusion, and blood protein digested in the gut is absorbed and metabolised to urea. A markedly raised urea in this setting is therefore a fairly specific marker of a significant upper GI bleed, which is why it can contribute up to 6 of the 23 points.
Can the Glasgow-Blatchford score be used for variceal bleeding?#
It can be calculated, and known hepatic disease scores 2 points, but the score does not distinguish variceal from non-variceal bleeding and was not designed to. Variceal bleeding carries higher mortality and needs vasoactive drugs and antibiotics started before endoscopy, so suspected varices should drive management regardless of the score.
References
Original / primary reference
Comparison
- Comparison of the Glasgow-Blatchford and Rockall Scores for prediction of nonvariceal upper gastrointestinal bleeding outcomes in Chinese patients. Medicine (Baltimore). 2019;98(21):e15716.
- Comparison of Glasgow-Blatchford Score and Rockall Score in Patients with Upper Gastrointestinal Bleeding. 2020.