Rockall Score for Upper GI Bleeding (Pre-Endoscopy)

Calculate the pre-endoscopy Rockall Score to stratify mortality and rebleeding risk in acute upper gastrointestinal haemorrhage using age, shock status, and comorbidity. Validated in over 5,000 patients.

Pre-Endoscopy Rockall Score Calculator

Select the appropriate option for each clinical variable. The pre-endoscopy (clinical) Rockall Score uses age, shock status, and comorbidity to predict mortality and rebleeding risk in acute upper GI haemorrhage before endoscopy is performed.

Age

Shock

Comorbidity

Disclaimer: The pre-endoscopy Rockall Score is an educational clinical decision-support tool. It does not replace clinical judgement, comprehensive patient assessment, or institutional protocols. The score should be used in conjunction with the full clinical picture, including laboratory results, medication history, and suspected aetiology. Always follow local guidelines for the management of acute upper gastrointestinal haemorrhage.

Pre-Endoscopy Rockall Score: formula, variables, and clinical context

Overview

The Rockall Score is a clinical risk-scoring system developed to predict mortality and rebleeding risk in patients presenting with acute upper gastrointestinal (GI) haemorrhage. The pre-endoscopy (clinical) version uses three variables obtainable at presentation: patient age, haemodynamic shock status, and comorbidity. It was derived using multiple logistic regression on a cohort of 4,185 patients identified during a national UK audit of acute upper GI haemorrhage in 1993, with subsequent validation on a prospective cohort of 1,625 patients.

Scoring variables

Variable	Criteria	Points
Age	< 60 years	0
	60-79 years	+1
	≥ 80 years	+2
Shock	No shock (HR < 100, SBP ≥ 100 mmHg)	0
	Tachycardia (HR ≥ 100, SBP ≥ 100 mmHg)	+1
	Hypotension (SBP < 100 mmHg)	+2
Comorbidity	No major comorbidity	0
	CHF, IHD, or any major comorbidity	+2
	Renal failure, liver failure, or disseminated malignancy	+3
Total score range		0-7

Risk stratification

Score	Risk category	Rebleeding	Mortality
0	Low risk	~0%	~0%
1-2	Intermediate risk	~0%	~0.2%
3-4	High risk	7-21%	0-10%
5-7	Very high risk	75-100%	25-100%

Risk estimates based on Tham et al. validation study (n = 102) of the clinical Rockall Score in acute non-variceal upper GI haemorrhage.

Outcomes by clinical Rockall Score (Tham et al.)

Score	Patients	Transfusion	Rebleeding	Mortality
0	38 (37%)	0%	0%	0%
1	13 (13%)	15%	0%	0%
2	16 (16%)	25%	0%	0%
3	16 (16%)	44%	0%	0%
4	14 (14%)	50%	7%	0%
5	4 (4%)	50%	75%	25%
7	1 (1%)	100%	100%	100%

Data from Tham TCK, James C, Kelly M. Predicting outcome of acute non-variceal upper gastrointestinal haemorrhage without endoscopy using the clinical Rockall Score. Postgrad Med J. 2006;82(973):757-759.

Variable definitions

Age (0-2 points)

Patient age at the time of presentation with acute upper GI haemorrhage. Advanced age is an independent predictor of mortality in UGIB. Patients under 60 receive 0 points, those aged 60-79 receive 1 point, and those 80 years or older receive 2 points. The original study found that age was one of the strongest independent predictors of mortality, reflecting both reduced physiological reserve and higher prevalence of comorbid disease.

Shock (0-2 points)

Haemodynamic status at presentation, assessed by heart rate (HR) and systolic blood pressure (SBP). "No shock" is defined as HR < 100 bpm and SBP ≥ 100 mmHg (0 points). Tachycardia alone (HR ≥ 100, SBP ≥ 100) scores 1 point, while hypotension (SBP < 100, regardless of HR) scores 2 points. Haemodynamic compromise reflects the severity of acute blood loss and is a critical determinant of short-term prognosis.

Comorbidity (0, 2, or 3 points)

The presence and severity of concurrent disease. No major comorbidity scores 0 points. Cardiac failure, ischaemic heart disease (IHD), or any other major comorbidity scores 2 points. Renal failure, liver failure, or disseminated malignancy scores 3 points. Note the non-linear scoring: the jump from 0 to 2 reflects the substantial prognostic impact of even moderate comorbid disease, while the organ failure/malignancy category carries the highest weighting in the entire score. "Major comorbidity" includes conditions such as COPD, diabetes, hypertension, cerebrovascular disease, and haematological malignancy.

Pre-endoscopy vs. complete Rockall Score

The complete Rockall Score incorporates two additional endoscopic variables: endoscopic diagnosis (0-2 points) and stigmata of recent haemorrhage (0-2 points), for a total range of 0-11. The pre-endoscopy (clinical) version (0-7) is designed to be calculated at the time of presentation, before endoscopy is available, enabling early risk stratification and triage.

The clinical Rockall Score is more conservative than the complete score in identifying low-risk patients (37% vs. 30% in validation studies), providing a greater margin of safety when considering early discharge. A clinical Rockall Score of 0 has been shown to identify patients with essentially no adverse outcomes who may be suitable for outpatient management or elective endoscopy.

Study details

Design: National prospective audit of acute upper gastrointestinal haemorrhage across 74 hospitals in the National Health Service (UK), 1993.
Derivation cohort: 4,185 cases of acute upper GI haemorrhage over a 4-month period.
Validation cohort: 1,625 subsequent cases over a 3-month period.
Statistical method: Multiple logistic regression to identify independent predictors of mortality. Five variables (age, shock, comorbidity, endoscopic diagnosis, stigmata of recent haemorrhage) were selected from 16 candidate variables.
Overall mortality: 14% in the audit population (including both variceal and non-variceal causes).
Score 0 identification: The pre-endoscopy score identified approximately 15% of patients at presentation as low risk (score 0).

Limitations

Subjective comorbidity assessment: The definition of "major comorbidity" is somewhat broad and subjective, leaving room for inter-observer variability in scoring.
Does not incorporate laboratory data: Haemoglobin, blood urea nitrogen (BUN), and coagulation parameters are not included, unlike some competing scores (e.g., Glasgow-Blatchford Score).
Limited discrimination at intermediate scores: Scores 1-3 showed no rebleeding or mortality in the Tham et al. validation, but transfusion rates increased substantially. The score may not adequately differentiate within this intermediate range.
Original cohort characteristics: The derivation cohort was a UK population from 1993. Changes in endoscopic therapy, pharmacology (e.g., proton pump inhibitors), and supportive care since then may affect the applicability of the original risk estimates.
Does not predict need for intervention: The Glasgow-Blatchford Score may be better at predicting which patients need clinical intervention, whereas the Rockall Score focuses on mortality prediction.

Key references

Rockall TA, Logan RFA, Devlin HB, Northfield TC. Risk assessment after acute upper gastrointestinal haemorrhage. Gut. 1996;38(3):316-321.
Rockall TA, Logan RFA, Devlin HB, Northfield TC. Selection of patients for early discharge or outpatient care after acute upper gastrointestinal haemorrhage. Lancet. 1996;347:1138-1140.
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.
Tham TCK, James C, Kelly M. Predicting outcome of acute non-variceal upper gastrointestinal haemorrhage without endoscopy using the clinical Rockall Score. Postgrad Med J. 2006;82(973):757-759.
Gralnek IM, Dulai GS. Incremental value of upper endoscopy for triage of patients with acute non-variceal upper GI haemorrhage. Gastrointest Endosc. 2004;60:9-14.

Practice caveats

The pre-endoscopy Rockall Score was validated primarily in non-variceal upper GI bleeding. It may not be applicable to patients with suspected variceal haemorrhage (e.g., known cirrhosis, signs of portal hypertension).
A clinical Rockall Score of 0 identifies a low-risk group suitable for potential early discharge or outpatient endoscopy, but this should be considered alongside social circumstances, access to follow-up, and local resource availability.
The Glasgow-Blatchford Score (GBS) is an alternative pre-endoscopy risk score that incorporates laboratory values and may be better at identifying patients who need clinical intervention, whereas the Rockall Score focuses primarily on mortality prediction.
Always integrate scoring tools with clinical judgement, medication history (e.g., anticoagulants, NSAIDs), and the full clinical presentation when making management decisions.

Background: The Burden of Upper Gastrointestinal Haemorrhage

Acute upper gastrointestinal (GI) haemorrhage remains one of the most common gastroenterological emergencies, with an annual incidence of approximately 50 to 150 per 100,000 population in developed countries. It is defined as bleeding originating proximal to the ligament of Treitz and encompasses a wide spectrum of presentations, from minor self-limited mucosal oozing to catastrophic arterial haemorrhage. The clinical manifestations include haematemesis (vomiting of fresh blood or coffee-ground material), melaena (black, tarry stools), and, in cases of brisk haemorrhage, haematochezia (passage of fresh blood per rectum despite an upper GI source).

Despite substantial advances in endoscopic haemostasis, pharmacotherapy (particularly intravenous proton pump inhibitors), interventional radiology, and critical care support, the overall mortality from acute upper GI bleeding has remained stubbornly between 6% and 14% across contemporary series. Much of this mortality is attributable not to exsanguination itself but to the interaction between acute haemorrhage and the patient's underlying physiological reserve. Elderly patients with significant comorbidities tolerate blood loss poorly, and it is frequently the decompensation of pre-existing cardiac, hepatic, or renal disease, rather than the bleed per se, that proves fatal.

This heterogeneity in outcomes creates a fundamental clinical dilemma at the point of presentation. A young, otherwise healthy patient with a small Mallory-Weiss tear has an entirely different trajectory from an 82-year-old with cirrhotic liver disease and variceal haemorrhage. Both present as "upper GI bleeding," but their management needs, resource requirements, and prognoses are worlds apart. The central challenge, therefore, is rapid and reliable risk stratification: identifying which patients require urgent intervention and intensive monitoring and which can be safely managed in a lower-acuity setting or even discharged for outpatient endoscopy.

The National UK Audit: Origins of the Rockall Score

The Rockall Score was born from one of the largest prospective studies of upper GI haemorrhage ever conducted. In 1993, Rockall, Logan, Devlin, and Northfield organised a national audit across 74 hospitals within the UK National Health Service. Over a four-month period, 4,185 consecutive cases of acute upper GI haemorrhage were identified and systematically recorded. The scope of the audit was remarkable for its era: it captured data on patient demographics, presenting features, haemodynamic status, comorbidities, endoscopic findings, therapeutic interventions, and clinical outcomes including rebleeding, the need for surgery, and death.

The overall mortality in the audit population was 14%, a figure that reflected the unselected nature of the cohort (including both variceal and non-variceal bleeding, emergencies, and inpatient bleeds). Using multiple logistic regression analysis, the investigators identified five independent predictors of mortality from an initial pool of 16 candidate variables. These five variables were: patient age, haemodynamic shock status (heart rate and systolic blood pressure), comorbidity, endoscopic diagnosis, and endoscopic stigmata of recent haemorrhage. Each variable was assigned a point value proportional to its regression coefficient, and the sum of points constituted the Rockall Score.

Critically, the investigators recognised that three of these five variables (age, shock, comorbidity) were available at the time of presentation, before any endoscopic procedure. This allowed the creation of two versions of the score: the pre-endoscopy (clinical) Rockall Score, which uses only the three clinical variables (score range 0 to 7), and the complete (post-endoscopy) Rockall Score, which incorporates all five variables (score range 0 to 11). The pre-endoscopy version was specifically designed to enable risk stratification at the point of first clinical contact, before endoscopy resources are committed.

The Pre-Endoscopy Rockall Score: Components

The pre-endoscopy Rockall Score is calculated from three variables, each assessed at the time of presentation.

Age (0 to 2 Points)

Patient age at the time of presentation with upper GI haemorrhage is categorised into three bands: younger than 60 years (0 points), 60 to 79 years (1 point), and 80 years or older (2 points). Age was one of the strongest independent predictors of mortality in the original logistic regression model. The relationship between age and outcome in upper GI bleeding is multifactorial. Older patients have diminished physiological reserve, reduced cardiovascular compensation for acute blood loss, higher prevalence of comorbid diseases, greater use of anticoagulant and antiplatelet medications that impair haemostasis, and reduced tolerance for the haemodynamic stresses of resuscitation and endoscopic procedures. In the original audit, patients aged 80 years or older had a mortality rate approximately three times that of patients younger than 60.

Shock (0 to 2 Points)

Haemodynamic status at presentation is categorised using heart rate (HR) and systolic blood pressure (SBP). Three levels are defined: no shock (HR below 100 beats per minute and SBP 100 mmHg or higher, scoring 0 points), tachycardia (HR 100 or higher with SBP 100 or higher, scoring 1 point), and hypotension (SBP below 100 mmHg regardless of heart rate, scoring 2 points).

The haemodynamic response to acute blood loss follows a well-characterised physiological sequence. In the initial compensated phase, sympathetic activation produces tachycardia and peripheral vasoconstriction, maintaining blood pressure despite ongoing volume loss. As haemorrhage progresses, compensatory mechanisms are overwhelmed and hypotension develops, signalling a more advanced stage of haemodynamic compromise. The Rockall Score captures this progression: tachycardia with maintained blood pressure indicates moderate haemorrhage with intact compensation, while hypotension indicates more severe or decompensated haemorrhage.

It is important to recognise that certain medications can blunt or obscure these haemodynamic responses. Beta-blockers suppress the tachycardic response, potentially masking significant blood loss behind a falsely reassuring heart rate. Patients on antihypertensive therapy may present with a "normal" systolic blood pressure of 100 mmHg that actually represents a significant drop from their baseline of 160 mmHg. Clinicians should interpret the shock variable in the context of the patient's medication history and baseline vital signs.

Comorbidity (0, 2, or 3 Points)

The comorbidity component uses a three-tier classification that assigns disproportionately heavy weight to concurrent disease. No major comorbidity scores 0 points. Cardiac failure, ischaemic heart disease (IHD), or any other major comorbidity scores 2 points. Renal failure, liver failure, or disseminated malignancy scores 3 points.

The non-linear weighting is deliberate and clinically significant. There is no 1-point category for comorbidity; the score jumps directly from 0 to 2 when any major comorbidity is present. This reflects the original regression analysis, which showed that even moderate comorbid disease had a substantial independent effect on mortality after upper GI bleeding. The 3-point category for organ failure and disseminated malignancy represents the highest single-variable contribution in the entire pre-endoscopy score, reflecting the profound prognostic impact of these conditions.

The definition of "major comorbidity" in the 2-point tier is intentionally broad and includes conditions such as chronic obstructive pulmonary disease (COPD), diabetes mellitus, hypertension with end-organ damage, cerebrovascular disease, peripheral vascular disease, haematological malignancy, and other systemic illnesses that reduce physiological reserve. Cardiac failure and IHD are specifically named because of their particularly strong association with adverse outcomes in the setting of acute haemorrhage, where the additional oxygen delivery demands of anaemia place the compromised myocardium at risk of ischaemia and decompensation.

The 3-point tier (renal failure, liver failure, disseminated malignancy) identifies patients with the most limited physiological reserve. Renal failure impairs platelet function, fluid balance, and drug clearance. Liver failure causes coagulopathy, portal hypertension (with the attendant risk of variceal haemorrhage), and impaired hepatic synthesis of clotting factors. Disseminated malignancy reflects a catabolic, immunocompromised state with limited life expectancy, where the prognosis is driven more by the underlying disease than by the bleeding episode itself.

The Complete Rockall Score: Post-Endoscopy Variables

While this calculator implements the pre-endoscopy version, understanding the complete score provides important clinical context. After endoscopy, two additional variables are incorporated.

Endoscopic Diagnosis (0 to 2 Points)

The endoscopic finding that identifies the source of bleeding is categorised as follows: no lesion identified, or a Mallory-Weiss tear (0 points); peptic ulcer disease, erosive disease, or oesophagitis (1 point); and malignancy of the upper GI tract (2 points). The endoscopic diagnosis provides critical prognostic information. A Mallory-Weiss tear has an excellent prognosis with very low rebleeding risk, while an upper GI malignancy carries a much higher risk of persistent or recurrent haemorrhage and reflects the underlying poor prognosis of the neoplasm. Peptic ulcer disease occupies an intermediate position, with outcomes heavily dependent on the endoscopic appearance of the ulcer (addressed by the next variable).

Stigmata of Recent Haemorrhage (0 to 2 Points)

This variable captures the endoscopic appearance of the bleeding lesion, which is one of the most powerful predictors of rebleeding risk. The classification is: no stigmata, or a clean-based ulcer with a dark spot (0 points); blood in the upper GI tract, an adherent clot, a visible non-bleeding vessel, or active spurting haemorrhage (2 points). The Forrest classification of peptic ulcer haemorrhage provides the conceptual framework for this variable. Active arterial spurting (Forrest Ia) and a visible non-bleeding vessel (Forrest IIa) carry the highest rebleeding rates (up to 55% and 43%, respectively, without endoscopic therapy), while a clean-based ulcer (Forrest III) has a rebleeding rate below 5%.

Complete Score Range and Risk Stratification

The complete Rockall Score ranges from 0 to 11. In the original validation study of 1,625 patients, mortality increased progressively with score: 0% for scores 0 to 2, 2.4% for score 3, 5.3% for score 4, 10.8% for score 5, 17.3% for score 6, 27.0% for score 7, and 41.1% for scores 8 and above. The addition of the endoscopic variables improved the discrimination of the model compared to the pre-endoscopy version alone, particularly in the intermediate score range where endoscopic findings provide the most incremental prognostic information.

Risk Stratification Using the Pre-Endoscopy Score

The pre-endoscopy Rockall Score divides patients into risk categories based on the total score at presentation.

Score	Risk Category	Rebleeding	Mortality	Management Implication
0	Low risk	~0%	~0%	Potential early discharge or outpatient endoscopy
1-2	Intermediate risk	~0%	~0.2%	Admission with endoscopy within 24 hours
3-4	High risk	7-21%	0-10%	Admission, close monitoring, endoscopy within 24 hours
5-7	Very high risk	75-100%	25-100%	ICU or HDU, urgent endoscopy, consider surgical/IR standby

The Low-Risk Group: Clinical Rockall Score of Zero

The most clinically actionable finding from the Rockall Score is the identification of the low-risk group with a score of 0. These are patients younger than 60 years, with no haemodynamic compromise (HR below 100, SBP 100 or above), and no major comorbidity. In both the original validation and subsequent studies, patients with a clinical Rockall Score of 0 had no rebleeding, no need for surgical intervention, and no mortality.

In the Tham et al. validation study of 102 patients with non-variceal upper GI bleeding, 38 patients (37%) had a clinical Rockall Score of 0. None of these patients required blood transfusion, none rebled, and none died. This finding has important resource implications: more than a third of patients presenting with upper GI bleeding may be safely identified for early discharge or outpatient management at the point of presentation, before endoscopy is performed.

The concept of early discharge for low-risk patients has been endorsed by multiple international guidelines. The UK National Institute for Health and Care Excellence (NICE) guidelines on acute upper GI bleeding recommend that patients with a pre-endoscopy Rockall Score of 0 be considered for early discharge with outpatient endoscopy. Similarly, international consensus statements have supported risk-score-guided triage to reduce unnecessary hospital admissions for low-risk upper GI bleeding.

The Intermediate and High-Risk Groups

Patients with scores of 1 to 2 represent an intermediate-risk group. In validation studies, these patients had no rebleeding or mortality but increasingly required blood transfusion (15% at score 1, 25% at score 2). These patients should be admitted for inpatient endoscopy within 24 hours, with routine haemodynamic monitoring and serial haemoglobin measurements.

Scores of 3 to 4 identify high-risk patients with substantial transfusion requirements (44% to 50%) and measurable rebleeding risk (7% at score 4 in the Tham cohort). These patients require close monitoring, adequate intravenous access, group-and-save or crossmatch, and timely endoscopy. Early involvement of the gastroenterology team is essential.

The Very High-Risk Group

Scores of 5 to 7 identify patients at extreme risk. In the Tham validation, patients with a score of 5 had a 75% rebleeding rate and 25% mortality, while the single patient with a score of 7 had 100% rates across all adverse outcomes. These patients require immediate resuscitation, consideration of intensive care or high-dependency unit admission, urgent endoscopy, and early multidisciplinary planning including surgical and interventional radiology standby. Massive transfusion protocols should be activated when clinically indicated.

Validation Studies and External Performance

The original Rockall Score was validated internally on a prospective cohort of 1,625 patients from the same national audit. Subsequent external validations have been conducted across multiple countries and healthcare settings.

Vreeburg et al. (1999) validated the complete Rockall Score in a Dutch cohort of 1,443 patients and found that the score effectively discriminated between low- and high-risk patients for both mortality and rebleeding, although the area under the receiver operating characteristic curve (AUROC) was moderate (0.67 to 0.70 for mortality prediction with the complete score). The pre-endoscopy version performed slightly less well in discrimination but retained its principal value: identifying a low-risk group with excellent outcomes.

Multiple other validation studies from Portugal, Korea, Malaysia, and other countries have generally confirmed the score's ability to identify a low-risk group with near-zero adverse outcomes when the clinical Rockall Score is 0. Performance in the intermediate range (scores 2 to 4) has been more variable, and the score's ability to discriminate between patients who will and will not rebleed has been consistently modest.

A key finding across validation studies is that the pre-endoscopy Rockall Score tends to be more conservative than the complete score in classifying patients as low risk. Approximately 15% to 37% of patients achieve a clinical Rockall Score of 0, compared to 20% to 30% who achieve a complete Rockall Score of 0 to 2. This conservatism provides an additional margin of safety when the pre-endoscopy score is used to guide early discharge decisions before the endoscopic diagnosis is known.

Comparison with the Glasgow-Blatchford Score

The Glasgow-Blatchford Score (GBS) is the principal alternative pre-endoscopy risk assessment tool for upper GI bleeding. Developed by Blatchford et al. in 2000 from a Scottish cohort, the GBS incorporates clinical variables (blood urea, haemoglobin, systolic blood pressure, heart rate, melaena, syncope, hepatic disease, cardiac failure) to predict the need for clinical intervention (transfusion, endoscopic therapy, surgery, or death). The GBS ranges from 0 to 23, with a score of 0 identifying patients at very low risk who can be safely discharged without endoscopy.

The fundamental difference between the Rockall and GBS scores lies in their design purpose. The Rockall Score was designed to predict mortality, while the GBS was designed to predict the need for intervention. This has practical implications for their optimal clinical roles.

Head-to-head comparisons have consistently shown that the GBS is superior at identifying patients who will require clinical intervention (transfusion, endoscopy, surgery). The GBS has higher sensitivity for the need for intervention and identifies a larger proportion of patients as truly low risk (GBS = 0). For example, the landmark study by Stanley et al. (2009, Lancet) compared the two scores in 676 patients and found that the GBS outperformed the pre-endoscopy Rockall Score for predicting the composite outcome of need for intervention (AUROC 0.90 vs 0.66).

However, the Rockall Score retains advantages in certain contexts. It is simpler to calculate (three variables vs eight in the GBS), requires no laboratory data (which may not be immediately available), and was specifically designed to predict the most clinically important outcome: mortality. In settings where rapid triage is needed before blood results return, the clinical Rockall Score can be calculated immediately at the bedside.

Current best practice, reflected in guidelines from NICE, the European Society of Gastrointestinal Endoscopy (ESGE), and the International Consensus Group, generally recommends the GBS as the preferred pre-endoscopy risk score for identifying patients suitable for outpatient management, with the Rockall Score serving as a complementary tool, particularly for mortality prediction and for the complete score after endoscopy has been performed.

The Role of Endoscopy Timing

The Rockall Score has implications for the timing of endoscopy. For patients with a clinical Rockall Score of 0, outpatient endoscopy (typically within 2 weeks) may be appropriate, as the risk of adverse outcomes during this waiting period is negligible. For patients with scores of 1 to 4, endoscopy within 24 hours of presentation is the standard recommendation per international guidelines. For patients with scores of 5 to 7, urgent endoscopy (within 12 hours or as soon as the patient is resuscitated) is indicated, given the very high risk of adverse outcomes.

The question of whether very early endoscopy (within 6 hours) improves outcomes compared to early endoscopy (within 24 hours) in high-risk patients has been addressed by several randomised trials, including the landmark 2020 trial by Lau et al. published in the New England Journal of Medicine. That trial found that urgent endoscopy within 6 hours did not reduce 30-day mortality compared to endoscopy within 6 to 24 hours in patients with acute upper GI bleeding and high-risk features (GBS 12 or higher). These findings suggest that while high-risk patients need endoscopy promptly, the priority should be haemodynamic resuscitation and stabilisation, with endoscopy performed within a 24-hour window rather than as an emergency procedure in an incompletely resuscitated patient.

Clinical Application: A Practical Workflow

In practice, the pre-endoscopy Rockall Score is calculated at the point of first assessment in the emergency department or acute medical unit. The workflow proceeds as follows:

Initial assessment and resuscitation: Secure airway, obtain intravenous access, begin fluid resuscitation, send blood for full blood count, urea and electrolytes, liver function tests, coagulation screen, and group-and-save. Administer a proton pump inhibitor (typically intravenous omeprazole or pantoprazole) if peptic ulcer bleeding is suspected.
Calculate the pre-endoscopy Rockall Score: Assess age, heart rate, systolic blood pressure, and comorbidity. Sum the points.
Triage based on score:
- Score 0: Consider early discharge with outpatient endoscopy within 2 weeks, provided the patient has reliable follow-up, no ongoing haemorrhage, and stable observations over a period of monitoring (typically 4 to 6 hours). Ensure the patient understands return precautions: recurrent haematemesis, melaena, presyncope, or worsening symptoms.
- Score 1-2: Admit for inpatient observation and endoscopy within 24 hours. Routine monitoring of vital signs and haemoglobin.
- Score 3-4: Admit to a monitored bed. Ensure crossmatched blood is available. Arrange endoscopy within 24 hours. Consider early gastroenterology consultation.
- Score 5-7: Admit to intensive care or high-dependency unit. Urgent endoscopy after initial resuscitation. Alert the surgical and interventional radiology teams. Activate massive transfusion protocol if indicated.
After endoscopy, calculate the complete Rockall Score: Add the endoscopic diagnosis and stigmata of recent haemorrhage components to refine the risk estimate. This informs decisions about the level of post-endoscopy monitoring, the duration of hospital stay, and the need for re-intervention planning.

The Rockall Score in Variceal Versus Non-Variceal Bleeding

The original Rockall Score derivation cohort included both variceal and non-variceal upper GI bleeding. However, the majority of validation studies, particularly the Tham et al. study from which the pre-endoscopy risk estimates are drawn, focused on non-variceal bleeding. Variceal haemorrhage has a distinct pathophysiology (portal hypertension, coagulopathy, liver synthetic failure), requires different therapeutic approaches (vasoactive drugs, band ligation, TIPS), and carries its own mortality prediction models (e.g., the Child-Pugh score, MELD score).

In patients with known or suspected variceal bleeding (known cirrhosis, stigmata of chronic liver disease, portal hypertensive gastropathy), the Rockall Score should be interpreted with caution. These patients will typically score at least 3 points on comorbidity alone (liver failure = 3 points), placing them automatically in the high-risk category. While this classification is directionally correct (variceal bleeding carries high mortality), the Rockall Score does not capture the specific prognostic variables relevant to variceal haemorrhage, such as the degree of liver synthetic failure, portal pressure, or the response to vasoactive therapy.

Anticoagulation and Antiplatelet Therapy

The Rockall Score was developed in an era before the widespread use of direct oral anticoagulants (DOACs: rivaroxaban, apixaban, edoxaban, dabigatran) and dual antiplatelet therapy (DAPT). These medications have substantially changed the epidemiology and management of upper GI bleeding. Patients on anticoagulants or antiplatelet agents may have more prolonged or difficult-to-control haemorrhage, may require reversal agents or haemostatic adjuncts, and face the additional challenge of balancing the risk of ongoing bleeding against the thromboembolic risk of stopping their medications.

The Rockall Score does not directly account for anticoagulation status. A patient on therapeutic warfarin with an INR of 4 and a patient with no anticoagulation receive the same Rockall Score if their age, haemodynamics, and comorbidities are identical. Clinicians should recognise this as a gap in the score and incorporate anticoagulation status into their overall risk assessment independently of the Rockall calculation. The comorbidity component may indirectly capture some of this risk, since conditions requiring anticoagulation (atrial fibrillation, mechanical heart valves, venous thromboembolism) often constitute "major comorbidity," but this is an imperfect proxy.

Transfusion Requirements as a Complementary Outcome

While the Rockall Score was designed to predict mortality and rebleeding, the Tham et al. validation data revealed a clear relationship between score and transfusion requirement. No patients with a score of 0 required transfusion, while 50% of patients with scores of 4 or 5 were transfused. This gradient provides additional clinical utility: the pre-endoscopy Rockall Score can help anticipate blood product requirements and guide preparedness.

Modern transfusion practice in upper GI bleeding has evolved considerably since the 1990s. The landmark TRIGGER and Villanueva et al. trials demonstrated that a restrictive transfusion strategy (transfusion threshold of haemoglobin 70 g/L) is associated with better outcomes than a liberal strategy (threshold 90 g/L) in most patients with upper GI bleeding, with the possible exception of patients with acute coronary syndromes. The Rockall Score does not incorporate haemoglobin level, which is a limitation: a patient with a haemoglobin of 55 g/L and a patient with 120 g/L receive the same Rockall Score if their clinical variables are identical. The GBS, which does include haemoglobin, may be more useful for predicting transfusion needs.

Limitations of the Rockall Score

Moderate Discrimination

The AUROC of the pre-endoscopy Rockall Score for predicting mortality is typically reported between 0.60 and 0.70, which represents modest discrimination. This means the score has a limited ability to rank individual patients by their true risk within the intermediate range (scores 1 to 4). Its greatest strength lies at the extremes: identifying the low-risk group (score 0) with near-zero adverse outcomes and the very-high-risk group (score 5 or above) with substantial mortality.

Subjective Comorbidity Classification

The definition of "major comorbidity" is deliberately broad, and there is room for inter-observer variability in deciding whether a given condition qualifies. Does well-controlled type 2 diabetes count as a "major comorbidity"? What about mild COPD? The original publication provided limited guidance on these borderline cases, and different clinicians may score the same patient differently. This subjectivity reduces the reproducibility of the score, particularly in the 0 versus 2 classification boundary that has the greatest clinical impact (since it determines whether the patient is in the low-risk or intermediate-risk group).

No Laboratory Component

The pre-endoscopy Rockall Score uses no laboratory values. While this is a deliberate design feature (enabling immediate bedside calculation), it means the score does not incorporate prognostically important information such as haemoglobin concentration, blood urea nitrogen (which rises with GI bleeding due to absorption of blood protein), international normalised ratio (INR), or lactate. The GBS addresses this limitation by including urea and haemoglobin, contributing to its superior performance for predicting the need for intervention.

Era of Derivation

The Rockall Score was derived from 1993 data. Since then, endoscopic haemostasis techniques have evolved substantially (epinephrine injection, thermal coagulation, mechanical clips, over-the-scope clips, haemostatic powders), intravenous PPI therapy has become standard, interventional radiology (transcatheter arterial embolisation) has expanded, and critical care has improved. These advances may have shifted the absolute mortality rates downward at each score level, potentially affecting the calibration (though not necessarily the discrimination) of the original risk estimates.

Does Not Predict Need for Intervention

The Rockall Score was designed to predict mortality, not the need for therapeutic intervention. A patient may have a low mortality risk but still require endoscopic haemostasis, transfusion, or hospitalisation. The GBS is generally considered superior for predicting the composite outcome of "need for clinical intervention," which may be a more clinically relevant endpoint for triage purposes than mortality prediction alone.

Special Populations and Considerations

Inpatient Upper GI Bleeding

Upper GI bleeding that develops in patients already hospitalised for another condition carries significantly higher mortality (up to 26% in some series) than community-onset bleeding. These patients typically have higher comorbidity burdens and less physiological reserve. The Rockall Score was derived from a mixed cohort that included inpatient bleeds, but most validation studies have focused on emergency department presentations. Clinicians should be aware that the Rockall risk estimates may underestimate mortality in inpatient-onset bleeding.

Elderly Patients

Patients aged 80 years or older constitute a growing proportion of upper GI bleeding presentations as populations age. These patients automatically receive 2 points for age, and the vast majority will have at least one major comorbidity (adding 2 or 3 points), placing them at a minimum score of 4 to 5 before haemodynamic status is even considered. While this accurately reflects their elevated risk, it also means the Rockall Score has limited granularity for discriminating within the elderly population, since nearly all elderly patients are classified as high or very high risk.

Patients on Proton Pump Inhibitor Therapy

A substantial proportion of patients presenting with upper GI bleeding are already taking proton pump inhibitors (PPIs) at the time of their bleed. PPI use at baseline may alter the endoscopic appearance of bleeding lesions (by reducing acid-related mucosal injury) and the spectrum of diagnoses encountered. The Rockall Score does not account for prior PPI use, and its impact on the score's calibration in contemporary PPI-prevalent populations has not been systematically studied.

Integration with Modern Risk Assessment Frameworks

Contemporary approaches to upper GI bleeding risk assessment increasingly use a complementary, multi-score strategy. The GBS (or its admission variant) is used at presentation to identify patients suitable for outpatient management (GBS = 0). For patients who are admitted and undergo endoscopy, the complete Rockall Score is calculated to refine the prognosis and guide post-endoscopy monitoring intensity and discharge timing.

The AIMS65 score (albumin below 30, INR above 1.5, altered mental status, systolic blood pressure 90 or below, age 65 or older) is another pre-endoscopy tool that has shown good performance for predicting inpatient mortality, particularly in US populations. Some centres use AIMS65 alongside or in place of the Rockall Score for mortality prediction.

The optimal strategy appears to be using the right tool for the right question: GBS for predicting who needs intervention and can be safely discharged, the pre-endoscopy Rockall or AIMS65 for mortality prediction, and the complete Rockall Score after endoscopy for refining prognosis and discharge planning. No single score captures all dimensions of risk in upper GI bleeding, and clinical judgement remains the essential integrating factor.