Prognosis

The incidence of GIB increases with age and comorbidities, as does the mortality associated. Multiple studies have shown that most patients do not die from blood loss but rather end-organ injury, decompensation of comorbidities, and complications from subsequent blood transfusion. In one study of more than 10,000 patients with peptic ulcer disease, 80% of patients died of non–bleeding-related causes; mainly multiorgan failure, cardiopulmonary disease, and terminal malignancy. Thus, prognosis correlates strongly with comorbidities. Identifying those patients at high risk for end-organ injury based on chronic medical conditions is imperative, specifically underlying pulmonary disease, coronary artery disease, cancer, liver disease, chronic alcoholism, and end-stage renal disease. Other historical features important for risk stratification include previous abdominal vascular surgery due to the risk of aortoenteric fistula development and the use of anticoagulants, including warfarin, novel anticoagulants, NSAIDs, and aspirin that independently are associated with worse outcomes. LGIB mortality also results from worsening comorbid conditions and nosocomial infection rather than severe bleeding. In one study of LGIB patients, advanced age, intestinal ischemia, and comorbid illness were the strongest predictors of mortality.

However, severe GIB, defined as documented GI blood loss with shock or a decrease in hematocrit level of 6%, does have a poor prognosis and is associated with mortality rates of 20% to 39%. In massive GIB with hemodynamic instability, UGIB is the most likely source. Predictors of severe UGIB include nasogastric aspirate with red blood, tachycardia, and hemoglobin levels below 8 mg/L. In LGIB, Strate and colleagues identified 7 factors that correlated with severe bleeding; heart rate greater than 100 beats per minute, systolic blood pressure less than 115 mm Hg, syncope, nontender abdominal examination, rectal bleeding within first 4 hours of evaluation, use of aspirin and greater than 2 comorbid illnesses. In patients with more than 3 of the 7 factors, the likelihood of severe bleeding was 80%.

Historical Features and Physical Examination Findings

Initially, the patient’s status can be determined by estimated blood volume loss and hemodynamic status. A patient’s reported blood loss is often inaccurate. It is helpful to assess the degree of blood loss with a functional assessment. Dizziness, lightheadedness, syncope, confusion, and weakness may be indicative of hypovolemia. Chest pain and more often dyspnea or weakness can be symptoms of myocardial ischemia in the setting of GIB. Bright red hematemesis suggests recent or active bleeding from the esophagus, stomach, or duodenum, whereas coffee ground emesis suggests bleeding that has resolved. Melena is an important finding of significant blood loss but can be seen with as little as 50 mL of blood.

Physical examination findings that correlate with low hemoglobin concentrations include color of the lower eyelid conjunctiva, nail bed rubor, nail bed blanching, and palmar crease rubor. Abnormal vital signs should be followed closely in patients with suspected GIB. Supine tachycardia is one of the most sensitive, early vital sign abnormalities indicating clinically significant blood loss. Orthostatics, in the setting of GIB, are recommended by multiple guidelines but its clinical value to identify hypovolemic patients remains controversial.

Laboratory Studies

Laboratory data can be helpful in the risk stratification and resuscitation of patients with GIB. A low initial hemoglobin less than 10 g/dL has been associated with higher mortality rates. However, a normal hemoglobin level can be falsely reassuring, as it can take up to 24 hours to accurately reflect the degree of blood loss. Coagulation profiles, platelet counts, and liver function tests can be helpful to assess for coagulopathy. As well, electrocardiogram and troponin should be liberally used to assess for coronary ischemia, as patients with GIB often experience ischemia without chest pain. In patients with known CAD, GIB is independently associated with higher mortality rates and ischemic complications.

Lactate has been well validated in trauma literature to aid clinicians in the assessment of acute blood loss. In critically ill patients admitted for GIB, El-Kersh and colleagues demonstrated that an admission lactate was predictive of outcome with a high sensitivity but low specificity. The median lactate level for nonsurvivors was 8.8 compared with 2.0 in survivors. Shah and colleagues found that inpatient mortality was 6.4-fold higher in those patients with an ED lactate greater than 4 mmol/L. In hemodynamically stable patients, point of care lactate elevation greater than 2.5 mmol/L was associated with hypotension within 24 hours with a 90% specificity and 84% negative predictive value and with a lactate greater than 5 mmol/L the specificity increased to 98% with a negative predictive value of 87%.

Prediction Scores

Multiple clinical prediction rules have been developed over the past couple of decades. In 1997, Kollef and colleagues in the original BLEED study developed a prediction rule to risk stratify patients in the emergency department with UGIB or LGIB as high or low risk of developing in-hospital complications of rebleeding, surgery or mortality. Kollef and colleagues found that visualized bleeding, hypotension, an elevated prothrombin time, “erratic” mental status, or unstable comorbid disease correlated with in-hospital complications and intensive care level care. Das and colleagues performed a derivation and validation study testing the original BLEED criteria with a triage simulation model to determine high risk patients that should be admitted to the ICU versus low risk patients that would be stable for the floor. Das and colleagues concluded that using the original BLEED criteria would result in unnecessary critical care admissions whereas a combination of visualized red blood by emesis or nasogastric aspirate and/or unstable comorbidities would more accurately assess which patients were at risk for decompensation in the first 24 hours of admission.

The 2 most commonly used clinical decision rules in the ED are the pre-endoscopic (Clinical) Rockall Score (CRS) and the Glasgow Blatchford Score (GBS). In 1996, Rockall and colleagues identified independent risk factors to accurately predict mortality. The scoring system uses clinical criteria (increasing age, comorbidity, shock) but disregards the endoscopic components (diagnosis, stigmata of acute bleeding) included in the full Rockall Score. In 2000, Blatchford and colleagues reported a prospective validation of a risk score based solely on clinical criteria that assesses the likelihood that a patient with an UGIB will need to have interventions, such as blood transfusion or endoscopy. Both GBS and CRS can accurately predict patients at risk for re-bleeding and mortality. However, GBS has been shown to be superior to the Rockall Score for predicting the need for admission, blood transfusion, or surgery. Furthermore, in one controlled study, a low-risk subset of patients with a GBS of 0 accounting for 16% of the study group were found to have no deaths or require interventions. The investigators concluded that these patients could be safely managed in an outpatient setting and subsequent studies extending the low-risk group to a GBS less than 1 concurs.

The AIMS65 score was developed and validated from a large database to predict inpatient mortality. AIMS65 is easy to use with a simple score calculation and well adapted for use in the emergency department. However, in a comparison study of AIMS65 to the Glasgow Blatchford Score, GBS was superior or equal to AIMS65 for all clinically relevant outcomes, including rebleeding, endoscopy, surgery, intensive care admission, and 30-day mortality. GBS was also superior to AIMS65 in identifying the lowest risk patients who might be appropriate for discharge.

See Table 1 for risk stratification scores.

Summary

Early identification of high-risk patients allows appropriate and timely interventions that may improve morbidity and mortality. Clinical predictors of increased risk for rebleeding or mortality include age older than 65 years, shock, poor overall health, comorbid illnesses, low initial hemoglobin levels, tachycardia, hypotension, melena, transfusion requirement, fresh red blood on rectal examination or in the emesis, blood on nasogastric aspirate, sepsis, elevated blood urea nitrogen (BUN), or elevated prothrombin time. Other factors predictive of poor outcomes include chronic alcoholism, active cancer, or poor social support conditions.

Examination of Stool

A close examination of the patient’s stool is naturally required for any patient with a suspected GIB. The presence of melena, for example, suggests an upper GI source of bleeding. With a demonstrated sensitivity of 80% in predicting an upper GI source, and a likelihood ratio of up to 5.9, ¹ the finding of melanotic stools is highly useful in localizing a GIB.

Although hemoccult-positive stool has often been used to detect LGIBs, positive results of this test do not completely rule-out an upper GI source. A study by Lee and colleagues, for example, found the guaiac-based test to be 17.9% sensitive at detecting an upper GI source.

Nasogastric Tube Aspiration

Controversy surrounds the placement of nasogastric tube (NGT) for aspiration in patients with suspected UGIBs, especially those without hematemesis. One retrospective study found that in patients admitted for GIBs without hematemesis, bloody nasogastric aspiration (NGA) had a positive likelihood ratio of 11 in detecting a UGIB. A negative NGA, however, had a likelihood ratio of 0.6, providing no useful diagnostic data. A recent systematic review also questioned the diagnostic utility for NGA, with sensitivity for upper GI bleeding ranging from 42% to 84%. Nonetheless, a positive NGA has been demonstrated to predict the presence of high-risk lesions at risk for rebleeding, with a 75.8% specificity.

Although a positive nasogastric aspiration is highly predictive of a UGIB, emergency physicians cannot rely solely on placement of a nasogastric tube to determine the need for emergent endoscopy. The clinician should discuss with the gastroenterologist how placement of the NGT will affect the timing in performing an upper endoscopy given that NGT placement is an uncomfortable and painful procedure for patients.

Blood Urea Nitrogen:Creatinine Ratio

An increase in BUN in many patients with UGIB has long been documented in the literature. More specifically, the ratio of BUN to creatinine (Cr) has been suggested as a means of differentiating UGIB from LGIB. In addition to decreased perfusion of the kidneys, it has been hypothesized that this finding may be due to the absorption of digested blood.

A systematic review found that a BUN:Cr ratio of greater than 30 is 93% specific for a UGIB, with a positive likelihood ratio of 7.5. Although the sensitivity of this method is relatively low (39% in a study by Witting and colleagues ), a positive test is highly predictive of an upper GI source in the setting of GIB.

Age

The prevalence of LGIB is directly associated with age. Indeed, a patient’s relative youth has been demonstrated to be an independent predictor for UGIB. One study found that age younger than 50 years to be 92% specific for an upper GI source, with a positive likelihood ratio of 3.5.