GI Bleeds: Who Needs ICU Level Care?

Authors: Christina Thorngren, MD, MPH and Janna Welch, MD (University of Texas Dell School of Medicine Emergency Medicine Residency Program) // Edited by: Erica Simon, DO, MHA (@E_M_Simon), Brit Long, MD (@long_brit ), and Alex Koyfman, MD (@EMHighAK, EM Attending Physician, UTSW / Parkland Memorial Hospital)

It’s a busy Saturday night in the ED. Scanning the patient tracking board, you come across an ESI 2: A 55 year-old woman with a chief complaint of generalized weakness. The nursing note details a review of systems positive for epigastric abdominal pain and black stools of one week duration. VS: HR 110, BP 105/64. As the patient rolls by in a wheelchair, you note her impressive pallor. After offering a quick greeting, placing her on the monitor, and dropping orders, you make your way to the clerk’s desk to notify him of an anticipated admission. As the clerk makes inquires regarding the required bed, your focus shifts to the patient: is she stable enough for the ward or does she require ICU level care?

If you’ve had a similar inner monologue while treating a GI bleed, the discussion below offers a number of tools to aid in your clinical decision-making.


Developing a clinical gestalt regarding a patient with a gastrointestinal bleed (GIB) can be challenging even for the seasoned emergency medicine physician. Anecdotally, we’ve all heard of the hemodynamically stable patient with one bloody bowel movement prior to arrival that acutely decompensates in the ED. While the decision to admit these patients to the ward versus the ICU may be clear in the setting of unstable VS or post endotracheal intubation, there are often times when we encounter shades of gray. The following discussion will hopefully shed some light on topic, and offer a quick discussion of risk stratification methods for EM physicians to utilize when addressing upper and lower GI bleeds.

Upper GIB

An upper GIB is defined as bleeding from a source proximal to the ligament of Treitz.1 Etiologies of upper GIBs include esophageal varices, peptic ulcers, gastritis, Mallory-Weiss tears, arteriovenous malformations, and rarely, Dieulafoy lesions (large diameter, tortuous vessels, protruding through the submucosa of the GI tract).2 Collectively, the annual incidence of upper GI bleeds is 48-180 cases per 100,000 adults, with a mortality ranging from 10-14%.1   Nearly 80% of upper GIBs resolve spontaneously, while 20% require acute intervention.1 Several studies have identified severe gastrointestinal bleeding (GI bleeding resulting in shock, or a decrease in hematocrit of ≥ 6% from baseline) as possessing a mortality rate of nearly 39%.1

While it is clear that patients with severe GI bleeds require inpatient admission, are there methods to determine when it is appropriate to discharge hemodynamically stable patients for outpatient follow-up?

Current literature cites the following scores for use in the mortality risk stratification of upper GIBs: the Clinical Rockall Risk Score,3 the Modified Glasgow-Blatchford Score,4,5 the AIMS65 Score,6 and the PNED Score.7,8 While these scores were initially developed to assess inpatient mortality in the setting of upper GIBs, secondary outcomes included risk for re-bleeds and 30 day mortality, making them useful tools for the emergency physician.3





Is one of these methods superior to the others? Let’s quickly discuss the statistical method of comparison:

The utility of a scoring systems is determined by calculating the area under the receiver operating curve (AUROC). An AUROC of 1 is a perfect test that when employed, will accurately and precisely predict an outcome 100% of the time. In contrast, an AUROC of 0.5 is of little utility as it precisely and accurately predicts an outcome 50% of the time (i.e. – no better than flipping a coin).

Ideally, to assess the risk stratification methods detailed above, all scores would be applied to one cohort of patients, and the AUROC calculated and compared. To date, there have been no studies performed which directly compare the Clinical Rockall Risk Score, the Modified Glasgow-Blatchford Score, the AIMS65 Score, and the PNED Score. Of the research published:

  • In 2012, Cheng, et al.5 compared the Modified Glasgow-Blatchford Score and the Clinical Rockall Score as applied to 167 patients presenting with GIBs (study endpoint: in-hospital mortality or re-bleeding). An AUROCs of 0.85 (CI 0.72-0.98) for the Modified Glasgow-Blatchford Score, and 0.59 (CI 0.32-0.87) for the Clinical Rockall Risk Score (p <0.0022) was identified.
  • Marmo, et al.8performed a head-to-head comparison of the PNED score and Clinical Rockall Score in 1360 patients (end point: in-hospital mortality) and found respective AUROCs of 0.81 (CI 0.70-0.90) for PNED, and 0.66 (CI 0.6-0.72) for the Clinical Rockall Score (p-value of <0.000).
  • In 2016, Aubougergi et al.6 performed a comparison of the AIMS65 and the Modified Glasgow-Blatchford Score in 298 patients (endpoint: inpatient mortality), identifying an AUROC of 0.85 (CI 0.81-0.89) for AIMS65 and 0.66 (CI 0.61-0.72) for Modified Glasgow-Blatchford Score (p of <0.01).

While it is unclear which mortality stratification method is most appropriate for use by the emergency physician, it is safe to say that the higher the mortality risk as characterized by these scores, the greater the necessity for advanced levels of patient care.

Have any of these scores been directly assessed for utility in predicting the need for ICU admission?

Of the mortality risk stratification scores above, only the Clinical Rockall Score has been evaluated for its utility in determining the requirement for ICU-level care. In a study of 565 consecutive patients treated for acute upper GIBs at Wellington Hospital, New Zealand (1988-1991), Phang et al.7 identified an overall mortality rate of 22% in patients presenting with a Clinical Rockall score of 4-7, leading the authors to identify this as a high-risk population requiring ICU level care.7

Publications to watch:

Of note, an abstract by Raemakers et al. was recently published online (prior to the full article in Academic Emergency Medicine), discussing the value of pre-endoscopic risk scores for upper GIBs in the ED. For more information as it becomes available:

Ramaekers R, Mukarram M, Smith C, Thiruganasambandamoorthy V. The predictive value of pre-endoscopic risk scores to predict adverse outcomes in emergency department patients with upper gastrointestinal bleeding — A systematic review. Acad Emerg Med 2016 Sep 19; [e-pub]. (

 Lower GI Bleeding

The majority of life-threatening bleeds originate from the upper GI tract, however profuse bleeding from the lower GI tract often causes hemodynamic instability. Approximately 20-25% of GIBs are distal to the ligament of Treitz, and result in a mortality rate of 2%- 4% (mortality has been demonstrated to increase with advancing age).9 Potential sources of lower GIBs include diverticular disease, malignancy, angiodysplasia, and colitis.9

Unlike upper GIBs, there is far less consensus regarding risk stratification parameters for lower GIBs. Several studies have demonstrated low systolic blood pressure, tachycardia, the use of aspirin, and the presence of medical comorbidities as increasing the risk of mortality in the setting of a lower GIB.9-11

In their study of 688 patients presenting with lower GIBs, Chong, et al.10 demonstrated a lack of reported abdominal tenderness as an independent risk factor for mortality (possibly as pain is often associated with more benign etiologies of bleeding, and could lead to earlier patient evaluation and treatment).10 The authors also identified prolonged bleeding (> 4 hours) as a risk factor for mortality in the setting of lower GIB.10,11

In conducting their study, Chong et al. also utilized a cohort of 410 patients to develop a clinical prediction score (HAKA score) to identify patients most likely presenting with a severe lower GIB. The authors identified severe bleeding as: bleeding requiring transfusion of ≥ 2 units of packed red blood cells, manifesting as a decrease in hematocrit of > 20% from baseline, recurrent bleeding within 24 hours, or readmission for lower GI bleeding within one week of initial presentation. The HAKA score, detailed below, demonstrated a PPV (for scores ≥ 2) of 44% for severe lower GI bleed, and a NPV of 88%.10


Recognizing the need to investigate risk factors for severe lower GIBs, in 2003 Strate and colleagues published data characterizing presenting symptoms of lower GIBs and their odds of association with severe bleeding (as defined above by Chong, et al.).11 In 252 consecutive patients presenting to Brigham and Womens’ hospitals in Boston, MA from 1996-1999, the following characteristics were associated with severe lower GI bleeding:



Extensive research has been performed in an attempt to develop clinical decision-making tools for the risk stratification of patients with GI bleeds. Ultimately, patients who are hemodynamically unstable, risk stratify as having a high mortality secondary to GI bleeding, or are at risk for having a severe lower GI bleed, should be admitted to an ICU setting.

Back to the Case

Review of the patient’s electronic health record reveals a history of chronic back pain (prescribed naproxen), hypertension, and hyperlipidemia. Initial screening labs identify a Hgb of 8.1 and BUN of 45. The patient is likely experiencing an upper GIB secondary to NSAID therapy. Utilizing the Modified Glasgow-Blatchford Scale, you quickly identify the patient as having a high risk of mortality. After consulting for ICU admission, you contact gastroenterology. The next morning you open the patient’s record and note the identification of a NSAID gastropathy on endoscopy. The patient ultimately required a blood transfusion, but is hemodynamically stable.

References / Further Reading

  1. Barkun A, Bardou M, Kuipers E, Sung J, Hunt R et al. International concensus recommendations on the management of patients with nonvariceal upper gastrointestinal bleeding. Ann Intern Med. 2010; 152:101-113.
  2. Baxter M and Aly E. Dieulafoy’s lesion: Current trends in diagnosis and management. Ann R Coll Surg Engl. 2010; 92(7): 548-554.
  3. Monteiro S, Goncalves T, Magalhaes J, Cotter J. Upper gastrointestinal bleeding risk scores: Who, when and why? World J Gastrointest Pathophysiol. 2016; 7(1):86-96.
  4. Blatchford O, Murray W, Blatchford M. A risk score to predict need for treatment for upper-gastrointestinal haemorrhage. Lancet. 2000; 356(9238):1318-1321.
  5. Cheng D, Lu Y, Sekhon H, Wu B. A modified glasgow blatchford score improves risk stratification in upper gastrointestinal bleed: a prospective comparison of scoring systems. Alimen Phamacol Ther. 2012; 36(8): 782-789.
  6. Abougergi M, Charpentier J, Berthea E, Rupawala A, Dheder J, et al. A prospective, multicenter study of the aims65 score compared with the Glasgow-blatchford score in predicting upper gastrointestinal hemorrhage outcomes. J Clin Gastroenterol. 2016; 50(6): 464-469.
  7. Phang T, Vornik V, Stubbs R. Risk assessment in upper gastrointestinal haemorrhage: implications for resource utilization. N Z Med J. 2000; 113(1115):331-333.
  8. Marmo R, Koch M, Cipolletta L, Capurso L, Grossi E, et al. Predicting moratlity in non-variceal upper gastrointestinal bleeders: validation of the Italian pned score and prospective comparison with the rockall score. Am J Gastroenterol. 2010;105(6):1284-1291.
  9. Qayed E, Dagar G, Nanchal R. Lower gastrointestinal hemorrhage. Crit Care Clin. 2016: 32(2):241-254.
  10. Chong V, Hill A, MacCormick A. Accurate triage of lower gastrointestinal bleed (LGIB) – a cohort study. Int J Surg. 2016; 25:19-23.
  11. Strate L, Orave E, Syngal S. Early predictors of severity in acute lower intestinal tract bleeding. Arch Intern Med. 2003;163(7):838-843.

Leave a Reply

Your email address will not be published. Required fields are marked *