Glasgow-Blatchford vs Rockall score in upper GI bleeding

What each score predicts

The Glasgow-Blatchford score and the Rockall score were built for different outcomes, which explains most of the difference in how they behave. Rockall and colleagues derived their score in Gut in 1996 from 4,185 cases of acute upper gastrointestinal haemorrhage across the UK in 1993, and tested it in 1,625 more in 1994, with death as the outcome. They found age, shock, comorbidity, endoscopic diagnosis, major stigmata of recent haemorrhage and rebleeding to be independent predictors of mortality. Blatchford, Murray and Blatchford derived their score in The Lancet in 2000 from 1,748 patients, with need for treatment as the outcome: blood transfusion, endoscopic or surgical intervention, rebleeding or death. So Rockall answers how likely a patient is to die, and Glasgow-Blatchford answers how likely they are to need a hospital intervention, which is the question that decides whether someone can go home. A patient can be low risk on one and not the other.

How the Rockall score is calculated

The Rockall score has five items: three clinical ones available at presentation and two that need an endoscopy. Age scores 0 under 60, 1 from 60 to 79 and 2 at 80 or over. Shock scores 0 with a pulse of 100 or less and systolic blood pressure of 100 mmHg or more, 1 for a pulse over 100 with a systolic pressure of 100 or more, and 2 for a systolic pressure under 100. Comorbidity scores 2 for ischaemic heart disease, cardiac failure or another major illness, and 3 for renal failure, liver failure or disseminated malignancy. These three give the pre-endoscopy (admission) Rockall score, 0 to 7. After endoscopy, diagnosis adds 0 for a Mallory-Weiss tear or no lesion, 1 for any other diagnosis and 2 for upper GI malignancy, and stigmata of recent haemorrhage add 2 for blood in the upper GI tract, an adherent clot or a visible or spurting vessel. The full Rockall score runs from 0 to 11.

Before or after endoscopy

The Glasgow-Blatchford score can be completed at the front door, while the full Rockall score has to wait for endoscopy. Glasgow-Blatchford uses only the admission hemoglobin and blood urea, pulse, systolic blood pressure, melena, syncope, and a history of liver disease or cardiac failure, all known within minutes of arrival and a set of bloods. That is why it is used for the decision the emergency department has to make first: whether the patient needs admission and urgent endoscopy at all. The admission Rockall score is also available early, but it was built to predict death, not need for treatment, and leaves out hemoglobin and urea; hemoglobin was among the factors not represented in Rockall's final model. In the original study, Rockall's score identified 15% of all cases at presentation, and 26% after endoscopy, as at low risk of rebleeding and negligible risk of death, possibly suitable for early discharge. The full Rockall score is mostly used after endoscopy, for prognosis and audit.

Which score performed better?

In the largest head-to-head study, the Glasgow-Blatchford score was clearly best at predicting intervention or death, while neither it nor Rockall was best at predicting death alone. Stanley and colleagues enrolled 3,012 consecutive patients at six hospitals on four continents (BMJ, 2017). For the composite of transfusion, endoscopic treatment, interventional radiology, surgery or 30-day death, the area under the ROC curve was 0.86 for Glasgow-Blatchford, against 0.70 for the full Rockall score, 0.66 for the admission Rockall score, 0.69 for PNED and 0.68 for AIMS65. A Glasgow-Blatchford score of 1 or less was the best threshold for survival without intervention, with a sensitivity of 98.6%. For 30-day mortality, AIMS65 and PNED did best at 0.77, the admission Rockall score reached 0.72, and Glasgow-Blatchford only 0.64. Optimal thresholds for predicting death were 4 or more on admission Rockall and 5 or more on full Rockall. No score predicted rebleeding or length of stay well.

Worked examples

Two patients show how the scores line up. A 72-year-old man with ischaemic heart disease presents with melena, a pulse of 105, systolic pressure of 105 mmHg, hemoglobin 11.0 g/dL and urea 9.0 mmol/L (BUN 25 mg/dL). Glasgow-Blatchford gives 3 for urea, 3 for hemoglobin, 1 for blood pressure, 1 for pulse and 1 for melena, a total of 9, above the 7 that best predicted endoscopic treatment in 2017. His admission Rockall score is 1 for age, 1 for shock and 2 for heart disease, a total of 4, the 2017 threshold for predicting death. If endoscopy then shows a duodenal ulcer with a visible vessel, his full Rockall score becomes 4 + 1 + 2 = 7. A 25-year-old woman with a small haematemesis after repeated vomiting, normal pulse and blood pressure, hemoglobin 13.0 g/dL and urea 5.0 mmol/L scores 0 on Glasgow-Blatchford and 0 on admission Rockall; a Mallory-Weiss tear with no stigmata keeps her full Rockall at 0.

Check the first patient in the Glasgow-Blatchford calculator, which accepts urea in mmol/L or BUN in mg/dL.