The Dispo of Patients with Upper GI Bleed

The goal for this JC was to discuss 2 commonly used scoring systems for patients with upper GI bleeds in an effort to identify patients who can be safely discharged from the ED. Both the Glasgow-Blatchford (GBS) and Rockall scores use clinical criteria, and the “full” Rockall also incorporates endoscopic findings.

Article 1: Stanley AJ, Ashley D, Dalton HR et al. Outpatient management of patients with low-risk upper-gastrointestinal haemorrhage: multicentre validation and prospective validation. Lancet. 2009;373:42-47.

This 2 part multi-center UK study used receiver-operating characteristic (ROC) curves to compare the ability of Glasgow-Blatchford (GBS) and Rockall scores (both pre and post-endoscopy Rockall) to predict need for clinical intervention (blood transfusion, endoscopic treatment, surgery) or death for 676 consecutive patients presenting with upper GI bleed. The area under the ROC curve was best for GBS (0.90 [95%CI 0.88-0.93), followed by full Rockall (0.81 [95% CI 0.77-0.84]), then pre-endoscopy Rockall (0.70 [95% CI 0.65-0.75]). In the second part of the study they prospectively identified 123 of 491 (22%) patients as low risk using a GBS of 0, of whom 84 (68%, or 15% of total cohort) were managed as outpatients, all without adverse events.

Bottom line article 1: A GBS of zero (BUN < 18, Hgb at least 13 for men/12 for women, SBP at least 110, Pulse < 100, absence of melena, syncope, CHF, liver dz) identifies a significant number of patients presenting with upper GI bleed who can safely be managed as outpatients.


  • Not a biggie, but Blatchford was one of the authors.
  • In phase 2, only 40% of patients underwent outpatient endoscopy. They did have 6 month follow-up through chart review and contact with patient/PMD and verified no readmissions for GI hemorrhage or death.
  • Louis-UK study, did not include demographic data on patients, and missing endoscopy on large percentage of patients treated as outpatients.
  • McKean-wanted to see information about age (concern about sending home old patients) and other demographics. Needs more external validation, especially with diverse patient groups.

Comments from the sages:

Harwood: As residents move forward in their careers, there will be an increasing emphasis on conserving resources and saving money in medicine. GBS could be a useful tool used in a dichotomous fashion (“home or admit”) to avoid unnecessary admissions, but needs more external validation.

Rogu: Likes GBS-it’s easy to use. Also, developing decision instruments is usually a multi-step process: derivation study followed by a validation study and then an impact analysis. This is a nice example of a study that both validates the GBS decision instrument and in part 2 examines its clinical impact.

Kulstad C: Think about GBS like PERC: it's helpful in avoiding a work-up, but a positive shouldn’t be followed blindly disregarding the specific clinical situation.


Article 2: Stephens JR, Hare NC, Warshow U et al. Management of minor upper gastrointestinal haemorrhage in the community using the Glasgow Blatchford Score. Eur J Gastroenterol Hepatol 2009;21:1340-1346.

In this two part UK study, GBS was first prospectively calculated on 232 patients admitted to a teaching hospital with upper GI bleed in order to identify a low risk cut-off score. All patients were offered next day or emergent endoscopy as indicated. GBS ≤2 and age <70 successfully predicted favorable  outcome (no need for endoscopic therapy, blood transfusion or surgery, or death) in all 52 of 232 (22%) patients with this score.

They then used this cut-off (GBS ≤2, age <70) to prospectively identify 104 of a second cohort of 304 patients (34%) as low risk, and considered outpatient management of these patients. Importantly, only 32 of these 104 (31% of low risk group, 10.5% of entire group) were actually managed as outpatients. Low risk patients were still admitted if there were concerns about social situation/support, active comorbidities, if taking warfarin or had suspected varices. All patients managed as outpatients were offered next weekday endoscopy, and nearly all completed endoscopy. All low risk patients in this cohort (both those admitted and those managed as outpatients) had good outcomes-no need for interventions, no deaths. In addition, 91% of patients managed as outpatients expressed a strong preference for this option.

Important-lack of need for acute intervention does not equal lack of findings on endoscopy- outpatient management still mandates availability of next working day endoscopy.

This is also a small study, and didn't include sensitivity/specificity/confidence intervals, and should be replicated with more patients and in other patient populations.

Bottom line for article 2: by using a GBS ≤2 with an amendment for age < 70, a set of low risk patients with upper GI bleed (10% % in this study) can be identified for consideration of safe outpatient management, but next-day endoscopy must be available.


  • As Erik often points out, there was no figure identifying how many potentially eligible patients were missed. If patients are enrolled M-F 9-5, the missed patients may constitute a different group with potentially different outcomes.
  • Janna: hard to set up endoscopy for all patients in our system. Liked that this approach reflects patient preference. Important observation in the study: rate of significant endoscopic findings dramatically increased with age (8% increase by decade of life).
  • Maletich: Even those admitted did well, although significant findings at endoscopy (18% vs. 9% findings in outpatient endoscopies)- reiterated challenge of outpatient endoscopy in US. This does help corroborate your decision to send some patients home.

Comments from the sages:

Harwood: Rockall score does not include hemoglobin. GBS includes hemoglobin. If one outcome measure is need for transfusion, then including hemoglobin introduces bias and favors your score.

Barbara: Treat seniors carefully!

Joan: NHS does not equal US health system-access to care and followup, availability of endoscopy much different.


Article 3: Chandra S, Hess EP, Agarwal D et al. External validation of the Glasgow-Blatchford Bleeding Core and the Rockall Score in the US setting. AJEM 2012;30:673-679.

In the last article, the authors attempt to externally validate the GBS and Rockall scores in the United States. In this retrospective study, 171 patients were identified with a diagnosis of upper GI Bleed. Primary outcome was need for endoscopic intervention, blood transfusion, angiography, surgery or death within 30 days. GBS outperformed pre-endoscopy Rockall (area under ROC curve 0.79 vs. 0.62), but was similar to post-endoscopy/full Rockall (area under ROC curve 0.72).

A GBS of 0 had a sensitivity of 99% (95% CI 93-100%) and specificity of 6% (95% CI 2-14%). For both the GBS and Rockall scores, at higher score cutoffs with greater specificity, the sensitivities were suboptimal.

Interestingly, 20 (12%) of cohort were discharged and managed as outpatients based on clinical gestalt, and only 2 required endoscopic intervention within 30 days (and no deaths), suggesting that EP decision making with appropriate consultation by GI is already safe and effective .

Bottom line for article 3: The prognostic accuracy of GBS and post-endoscopy Rockall were similarly high. Although the sensitivity of a GBS of 0 was excellent in this study, the low specificity of both GBS and Rockall prevents the authors from recommending either score as the sole means to identify patients safe for discharge/outpatient management.


  • Matt: Limitations of retrospective chart review. Also, study only included residents of Olmsted County, as all inpatient and outpatient medical records for these patients are available through a searchable data set. This excluded a large number of otherwise potentially eligible patients.
  • Jako: This retrospective relatively small study alone not strong enough to support use of GBS, and low specificity limits usefulness. Needs external validation beyond one county in Minnesota.

Comments from the sages:

Sam: Low specificity means higher number of false positives. Clinical judgment may perform better. Trying to explain the poorer fit of the ROC curves in this study compared to study 1, likely in part due to the methods of chart review, eg if patients with a small amount of bleeding are diagnosed with “gastritis” and discharged, they wouldn’t be picked up and counted, and it’s a group that likely would be GBS 0 with excellent outcomes.

Stats blurb: Receiver Operating Characteristic (ROC) curves.
Y axis is True Positive Rate = True positives/All positives = sensitivity. X axis is False Positive Rate = False positives/All negatives = 1 – specificity.

The area under the curve is a measure of text accuracy and shows the tradeoff between sensitivity and specificity. The closer the curve follows the left-hand border and then the top border of the ROC space, the more accurate the test. A perfect test with a sensitivity and specific of 100% would have an area under the curve of 1.0 In general, 0.9 is excellent, 0.8 is good, 0.7 is fair, and 0.5 is a straight line at 45 degrees which is worthless. ROC curves are also often used to determine an optimal cut off value for a test. The optimal cut-off point on the curve will vary depending on whether the clinical situation demands a high sensitivity or high specificity.

Journal club bottom line:
The Glasgow Blatchford Score is easy to use and a score of zero appears to help predict a low risk group of patients with upper GI bleed who can be considered for outpatient management. A score of GBS ≤2 and age <70 is another reasonable cut-off, and incorporates age. Outpatient management for a patient with a GBS of 0 is only a safe option if excellent follow-up and next day endoscopy is assured. Respect age as a predictor of more serious underlying pathology and risk for bad outcomes-nobody in the room was comfortable sending home elderly patients with GI bleeds. And while the GBS is useful, it should be an adjunct to your clinical judgment: remember to consider co-morbidities and the social situation of the patient.