About the Rockall Score for Upper Gastrointestinal Bleeding
Once endoscopy has been done for an acute upper gastrointestinal bleed, the Rockall score places mortality risk on a 0–11 scale: 0–2 is low risk, 3–5 intermediate, and 6 or above high risk. Five variables feed it — age, shock, comorbidity, endoscopic diagnosis, and stigmata of recent haemorrhage — and comorbidity carries the heaviest single weight at up to 3 points. Drop the two endoscopic items and what remains is the pre-endoscopy score out of 7, which is the version available at triage.
Formula
Rockall = age points + shock points + comorbidity points + endoscopic diagnosis points + stigmata points- age points
- 0, 1 or 2 — see thresholds above.
- shock points
- 0, 1 or 2, from heart rate and systolic blood pressure.
- comorbidity points
- 0, 2 or 3 — the heaviest-weighted single variable.
- endoscopic diagnosis points
- 0, 1 or 2, from the endoscopic finding.
- stigmata points
- 0 or 2, from endoscopic stigmata of recent haemorrhage.
- The pre-endoscopy score is the same formula with the last two (endoscopic) variables omitted, giving a maximum of 7 rather than 11.
- Comorbidity, not shock or age, carries the single largest per-item weight (up to 3 points) — a common source of underscoring when it is skipped as 'not applicable'.
- There is no continuous or log-transformed term anywhere in the score; every variable is an ordinal category assigned directly from a threshold or clinical category.
Interpreting the result
The score was built to answer a mortality question, and it should be read as one: it stratifies death risk well and rebleeding risk considerably less well, a distinction confirmed independently in external validation. A low score (0–2) supports early discharge or outpatient follow-up once endoscopic therapy, where needed, is complete. A high score (6 or above) should trigger a monitored bed and a pre-agreed escalation plan rather than a specific mandated intervention, since the score itself does not specify what to do next beyond flagging the level of risk.
| Score | Band | What it means | Action |
|---|---|---|---|
| 0–2 | Low risk | Low mortality and rebleeding risk in validation cohorts | Consider early discharge with outpatient follow-up once endoscopic management is complete |
| 3–5 | Intermediate risk | Moderate mortality risk | Admit for observation with a defined rebleeding escalation plan |
| 6–11 | High risk | Substantially higher mortality risk | High-dependency or intensive monitoring; plan for repeat endoscopy or surgery/interventional radiology before rebleeding occurs |
Scroll the table sideways for every column.
What the Rockall Score needs (5 inputs)
- Age
- 0 points under 60, 1 point at 60–79, 2 points at 80 or over. The single largest contributor to the pre-endoscopy subscore.
- Shock
- 0 points for no shock (heart rate under 100, systolic BP 100 or above), 1 point for tachycardia alone (heart rate 100 or above, systolic BP still 100 or above), 2 points for hypotension (systolic BP under 100, regardless of heart rate).
- Comorbidity
- 0 points for none of note, 2 points for cardiac failure, ischaemic heart disease or any other major comorbidity, 3 points for renal failure, liver failure or disseminated malignancy. This is the single highest-weighted variable in the score.
- Endoscopic diagnosis
- 0 points for a Mallory-Weiss tear or no lesion found, 1 point for peptic ulcer or other erosive disease, 2 points for upper GI malignancy. Requires endoscopy to score.
- Major stigmata of recent haemorrhage
- 0 points for none or a dark spot only, 2 points for blood in the upper GI tract, an adherent clot, or a visible or spurting vessel. Also requires endoscopy.
What it returns
- Full Rockall score (0–11)
- Computed once endoscopy is done. Sums all five variables.
- Pre-endoscopy score (0–7)
- Age, shock and comorbidity only — available at presentation, before the endoscopic findings exist.
- Risk band
- Low, intermediate, or high, mapped from the full score.
How it is calculated
The score comes from a multiple logistic regression fitted to a prospective, population-based UK audit of acute upper GI bleeding, and the point values are a simplified integer approximation of the regression coefficients — age, shock, comorbidity, diagnosis and major stigmata were the five variables that remained independently predictive of death, while haemoglobin, sex, mode of presentation, and NSAID or anticoagulant use did not add independent predictive value and were dropped from the final model. Because it is a sum of ordinal categories rather than a continuous equation, it discriminates well at the extremes (very low or very high risk) and less sharply through the middle of its range, which is typical of scores built this way.
Facts & figures
| Outcome | Better discriminator | Reported AUCs |
|---|---|---|
| Mortality | Rockall (pre-endoscopy and full) | 0.842 vs 0.622 (n=2,977, 8-centre China cohort); 0.648 vs 0.582 (n=182, Iran) |
| Need for blood transfusion | Glasgow-Blatchford | 0.757 vs 0.528 (n=182, Iran) |
| Rebleeding | Glasgow-Blatchford | 0.722 vs 0.520 (n=182, Iran) |
| Need for endoscopic intervention | Glasgow-Blatchford | 0.771 vs 0.650 (n=182, Iran) |
Scroll the table sideways for every column.
Both comparison cohorts agree on direction — Rockall wins mortality, Glasgow-Blatchford wins the pre-endoscopy triage questions — even though the absolute AUCs differ considerably between them, which is typical for this literature and worth stating rather than averaging away.
Evidence
Derivation — UK national audit
1996 · n = 4,185A prospective, population-based, two-phase UK audit run with the Association of Surgeons of Great Britain and Ireland and the British Society of Gastroenterology. Phase one enrolled 4,185 patients with acute upper GI haemorrhage over four months in 1993 to derive the score by multiple logistic regression; phase two enrolled 1,625 patients over three months in 1994 to test it in a separate population.
Age, shock, comorbidity, endoscopic diagnosis and major stigmata of recent haemorrhage were independent predictors of death on multivariable analysis; haemoglobin, sex, mode of presentation, NSAID use and anticoagulant use were not and were excluded from the final score. In the 1994 validation population the score reproducibly separated risk categories, identifying 26% of patients as low risk after endoscopy for possible early discharge.
External validation — Amsterdam cohort
1999Independent reanalysis comparing Rockall's own post-derivation validation sample with a separate external cohort of patients admitted to several Amsterdam hospitals (the 'Vreeburg' sample), assessing calibration and discrimination for both mortality and rebleeding.
Area under the ROC curve for mortality was 0.81 in Rockall's validation sample and 0.73 in the external Amsterdam sample. For rebleeding, discrimination was materially weaker in both: 0.70 and 0.61 respectively. Calibration was adequate for mortality in both samples but poor for rebleeding, confirming that the score's real strength is prognostic, not a rebleeding prediction tool.
Comparison with Glasgow-Blatchford — 8-centre China cohort
2019 · n = 2,977Retrospective multicentre cohort of 2,977 patients with non-variceal upper GI bleeding across eight Chinese hospitals, comparing pre-endoscopic Rockall, full Rockall and Glasgow-Blatchford for predicting in-hospital mortality.
AUC for mortality: pre-endoscopic Rockall 0.842, full Rockall 0.804, Glasgow-Blatchford 0.622 — both Rockall variants significantly outperformed Glasgow-Blatchford for this outcome (p<0.01).
Comparison with Glasgow-Blatchford — Iran cohort
2016 · n = 182Prospective single-centre cohort of 182 patients presenting with upper GI bleeding to an emergency department in Tehran, followed for rebleeding and one-month mortality.
Full Rockall outperformed Glasgow-Blatchford for mortality (AUC 0.648 vs 0.582, p=0.021), but Glasgow-Blatchford outperformed Rockall for transfusion need (0.757 vs 0.528), rebleeding (0.722 vs 0.520) and need for endoscopic intervention (0.771 vs 0.650, all p≤0.002).
How it compares
Rockall Score vs Glasgow-Blatchford Score
Use Glasgow-Blatchford at triage to decide who can avoid admission, and Rockall after endoscopy to gauge mortality risk — they answer different questions and neither substitutes for the other.
Across multiple comparison cohorts, Rockall (pre-endoscopy or full) discriminates mortality better than Glasgow-Blatchford, while Glasgow-Blatchford discriminates the need for transfusion, endoscopic intervention and rebleeding better than Rockall. Glasgow-Blatchford also needs no endoscopic findings, so it is available at first contact; Rockall's most useful form requires the scope to have already happened. The two scores were built for different decisions and the comparison literature bears that out rather than crowning one score outright better.
Rockall Score vs AIMS65
AIMS65 and pre-endoscopic Rockall serve the same triage moment and perform comparably for mortality; AIMS65 has the practical edge of being entirely lab- and mental-status-based, with no shock or comorbidity category to judge.
Both are usable before endoscopy and both are built primarily around mortality rather than intervention need, which distinguishes them from Glasgow-Blatchford. Neither has displaced the other in guidelines, and the choice in practice often comes down to which values are fastest to obtain in a given department.
Pearls & pitfalls
- Comorbidity carries up to 3 points, more than any other single item — do not treat it as an afterthought relative to shock or age.
- The score needs endoscopy to complete. Do not report a 'Rockall score' at triage; that is the pre-endoscopy subscore out of 7, and it is a different number with a different purpose.
- It predicts mortality much better than rebleeding — every independent validation has found the same pattern. Do not use a low Rockall score alone to reassure yourself that rebleeding is unlikely.
- It was derived in an unselected UGIB population, not a peptic-ulcer-only cohort, and it was not designed for variceal bleeding specifically.
- Shock is scored from the worse of heart rate or systolic BP category, not an average — a patient with a fast pulse but normal pressure still scores for tachycardia.
Critical actions
- Complete the pre-endoscopy subscore at presentation to help triage urgency of endoscopy, and the full score afterwards for mortality risk and discharge planning.
- Resuscitate and correct shock before over-weighting the endoscopic findings; the shock variable is scored at presentation, not after resuscitation.
- Do not let a low Rockall score substitute for a rebleeding management plan — use its own weaker rebleeding performance as a reason for continued clinical vigilance, not reassurance.
- Use Glasgow-Blatchford, not Rockall, if the question is whether a patient can be safely discharged before endoscopy.
- In a high-risk score, arrange a defined plan for repeat endoscopy, interventional radiology or surgery before rebleeding happens, not after.
Why this score exists
The score's original purpose was not the one it is mostly used for today. Rockall and colleagues built it from a prospective national audit of 4,185 cases, later tested on a further 1,625, and the paper states the intended application plainly: determining case mix when comparing outcomes between units in audit and research, and calculating risk-standardised mortality. Identifying individual low-risk patients was the secondary use, and the paper quantified it — 15% of cases at presentation and 26% after endoscopy carried a negligible risk of death. That audit origin explains the score's shape. It was fitted to predict mortality specifically, which is why it still outperforms Glasgow-Blatchford on death but loses to it on transfusion, rebleeding and need for intervention. It also explains the omissions: haemoglobin, sex, presenting features other than shock, and both NSAIDs and anticoagulants were all tested and none survived into the final model.
About the creator
First author, 1996 derivation study
Derived the score from a UK national audit of acute upper gastrointestinal haemorrhage, combining clinical and endoscopic variables to predict mortality.
Senior author
Co-authored the national audit steering group's analysis from which the score came.
Limitations
- Two of five variables require endoscopy, so the full score cannot be calculated at first presentation.
- Consistently the weakest performer of the common UGIB scores for predicting rebleeding specifically, across independent validations.
- Derived in an all-comer UGIB population; performance in variceal bleeding specifically has not been established to the same standard.
- Comorbidity is graded by category rather than a validated comorbidity index, which leaves some judgement in how a given condition is classified.
- Absolute AUCs for the same comparison (Rockall vs Glasgow-Blatchford) vary considerably between cohorts, so a single reported figure should be read as one data point rather than a fixed constant.
If you are the patient
The Rockall score is a number your medical team calculates after your endoscopy for a gastrointestinal bleed, using your age, whether you were in shock when you arrived, any other significant health conditions, and what the endoscopy found. It runs from 0 to 11, and a lower number means a lower estimated risk of dying from this bleed — it is mainly used to judge how closely you need to be monitored afterwards and whether you might be well enough to go home sooner. It is not a rebleeding forecast: studies have found it is much better at predicting the risk to life than at predicting whether the bleeding will start again, so your team will still watch you closely for signs of rebleeding regardless of the score.
Frequently asked questions
What is the Rockall score?#
A five-variable score for acute upper GI bleeding — age, shock, comorbidity, endoscopic diagnosis and stigmata of recent haemorrhage — that runs 0 to 11 and predicts mortality risk. It requires endoscopy to complete in full.
What is the difference between the pre-endoscopy and full Rockall score?#
The pre-endoscopy score uses only age, shock and comorbidity, for a maximum of 7, and can be calculated at first presentation. The full score adds the endoscopic diagnosis and stigmata of recent haemorrhage, for a maximum of 11, and needs the endoscopy to be complete.
What Rockall score is considered high risk?#
A score of 6 or above is generally treated as high risk, warranting intensive monitoring and a clear plan for further intervention if bleeding recurs. Scores of 0–2 are low risk and 3–5 intermediate.
Is Rockall or Glasgow-Blatchford score more accurate?#
It depends on the question. Rockall discriminates mortality better in the comparison literature; Glasgow-Blatchford discriminates the need for transfusion, endoscopic intervention and rebleeding better. Neither is uniformly superior — they were built to answer different clinical questions.
Does the Rockall score predict rebleeding?#
Poorly, and this has been shown consistently. External validation found an AUC of only 0.61–0.70 for rebleeding, compared with 0.73–0.81 for mortality. Use it mainly for mortality risk, not as a rebleeding forecast.
Can the Rockall score be used before endoscopy?#
Only its pre-endoscopy component — age, shock and comorbidity, maximum 7 — which is used to help gauge urgency. The full 0–11 score needs the endoscopic findings and cannot be completed beforehand.
References
Original / primary reference
Validation and comparison studies
- Vreeburg EM, Terwee CB, Snel P, et al. Validation of the Rockall risk scoring system in upper gastrointestinal bleeding. Gut. 1999;44(3):331-335.
- Lu M, Sun G, Huang H, et al. 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.
- Mokhtare M, Bozorgi V, Agah S, et al. Comparison of Glasgow-Blatchford score and full Rockall score systems to predict clinical outcomes in patients with upper gastrointestinal bleeding. Clin Exp Gastroenterol. 2016;9:337-343.