# 44 - 51 Gastrointestinal Bleeding

### 51 Gastrointestinal Bleeding

of UWL and is often more pronounced in men. More intense forms of 
mental illness such as paranoid disorders may lead to delusions about 
food and cause weight loss. Alcoholism can be a significant source of 
weight loss and malnutrition.

Elderly persons living in poverty may have to choose whether to 
purchase food or use the money for other expenses, including medica­
tions. Screening questions can probe whether patients have run out of 
food or whether they routinely purchase less than they need. Institu­
tionalization is an independent risk factor, as up to 30–50% of nursing 
home patients have inadequate food intake.
Medications can cause anorexia, nausea, vomiting, gastrointestinal 
distress, diarrhea, dry mouth, and changes in taste. This is particularly 
an issue in the elderly, many of whom take five or more medications.
PART 2
Cardinal Manifestations and Presentation of Diseases
■
■ASSESSMENT
The four major manifestations of UWL are (1) anorexia (loss of appetite), 
(2) sarcopenia (loss of muscle mass), (3) cachexia (a syndrome that 
combines weight loss, loss of muscle and adipose tissue, anorexia, and 
weakness), and (4) dehydration. The current obesity epidemic adds 
complexity, as excess adipose tissue can mask the development of 
sarcopenia and delay awareness of the development of cachexia. If it is 
not possible to measure weight directly, a change in clothing size, cor­
roboration of weight loss by a relative or friend, and a numeric estimate 
of weight loss provided by the patient are suggestive of true weight loss.
Initial assessment includes a comprehensive history and physical, a 
complete blood count, tests of liver enzyme levels, C-reactive protein, 
erythrocyte sedimentation rate, renal function studies, thyroid 
function tests, chest radiography, and an abdominal ultrasound 
(Table 50-2). Age-, sex-, and risk factor–specific cancer screening 
tests, such as fecal occult blood, colonoscopy, or mammography, 
should be performed (Chap. 75). Patients at risk should have HIV 
testing. All elderly patients with weight loss should undergo screen­
ing for dementia and depression by using instruments such as the 
Mini-Mental State Examination and the Geriatric Depression Scale, 
respectively (Chap. 489). The Mini Nutritional Assessment (www.
mna-elderly.com) and the Nutrition Screening Initiative (http://www.
ncbi.nlm.nih.gov/pmc/articles/PMC1694757/) are also available for the 
TABLE 50-2  Assessment and Testing for Involuntary Weight Loss
Indications
Laboratory
5% weight loss in 6 mo
Complete blood count
 
Comprehensive electrolyte and 
metabolic panel, including liver and 
renal function tests
Body mass index <21
Thyroid function tests
25% of food left uneaten after 7 d
Erythrocyte sedimentation rate
Change in fit of clothing
C-reactive protein
Change in appetite, smell, or taste
Ferritin
Abdominal pain, nausea, vomiting, 
diarrhea, constipation, dysphagia
HIV testing, if indicated
Assessment
Radiology
Complete physical examination, 
including dental evaluation
Chest x-ray
Abdominal ultrasound
Medication review
 
Recommended cancer screening
 
Mini-Mental State Examinationa
 
Mini-Nutritional Assessmenta
 
Nutrition Screening Initiativea
 
Simplified Nutritional Assessment 
Questionnairea
 
Observation of eatinga
 
Activities of daily livinga
 
Instrumental activities of daily livinga
 
aMay be more specific to assess weight loss in the elderly.

nutritional assessment of elderly patients. Almost all patients with a 
malignancy and >90% of those with other organic diseases have at 
least one laboratory abnormality. In patients presenting with substan­
tial UWL, major organic and malignant diseases are unlikely when 
a baseline evaluation is completely normal. Careful follow-up rather 
than additional undirected testing is advised because the prognosis of 
weight loss of undetermined cause is generally favorable.
TREATMENT
Unintentional Weight Loss
The first priority in managing weight loss is to identify and treat 
the underlying causes. Treatment of underlying metabolic, social, 
psychiatric, dental, infectious, or other systemic disorders may 
be sufficient to restore weight and functional status gradually. 
Medications that cause nausea or anorexia should be withdrawn or 
changed, if possible. For those with unexplained UWL, oral nutri­
tional supplements such as high-energy drinks sometimes reverse 
weight loss. Advising patients to choose appealing foods and to 
consume supplements between meals rather than with a meal may 
help minimize appetite suppression and facilitate increased over­
all intake. Orexigenic, anabolic, and anticytokine agents are not 
generally recommended. In selected patients, the antidepressant 
mirtazapine significantly increases body weight, body fat mass, and 
leptin concentration. However, side effects, including dizziness, 
fatigue, and somnolence, occur in about 10% of patients. Patients 
with wasting conditions who can comply with an appropriate exer­
cise program gain muscle protein mass, strength, and endurance 
and may be more capable of performing ADLs.
■
■FURTHER READING
Gaddey HL, Holder KK: Unintentional weight loss in older adults. 
Am Fam Physician 104:34, 2021.
McMinn J et al: Investigation and management of unintentional 
weight loss in older adults. BMJ 342:d1732, 2011.
Nicholson BD et al: Prioritising primary care patients with unexpected 
weight loss for cancer investigation. BMJ 370:m2651, 2020.
Perera LAM et al: Approach to patients with unintentional weight 
loss. Med Clin North Am 105:175, 2021.
Wong CJ: Involuntary weight loss. Med Clin North Am 98:625, 2014.
Loren Laine

Gastrointestinal 

Bleeding
Gastrointestinal bleeding (GIB) presents as either overt or occult 
bleeding. Overt GIB is manifested by hematemesis, vomitus of red 
blood or “coffee-grounds” material; melena, black, tarry stool; and/or 
hematochezia, passage of red or maroon blood from the rectum. In 
the absence of overt bleeding, occult GIB may present with symptoms 
of blood loss or anemia such as lightheadedness, syncope, angina, or 
dyspnea; with iron-deficiency anemia; or a positive fecal occult blood 
test on colorectal cancer screening. GIB is also categorized by the site of 
bleeding as upper, from the esophagus, stomach, or duodenum; lower, 
from the colon; small intestinal; or obscure GIB if the source is unclear.
GIB is the most common gastrointestinal condition leading to hos­
pitalization in the United States, accounting for ~530,000 admissions 
annually and a case fatality of ~2%. Patients generally die from decom­
pensation of other underlying illnesses rather than exsanguination.

■
■SOURCES OF GASTROINTESTINAL BLEEDING
Upper Gastrointestinal Sources of Bleeding 
• 
PEPTIC 
ULCERS  Peptic ulcers are the most common cause of upper GIB 
(UGIB), accounting for ~50% of UGIB hospitalizations. Features of 
an ulcer at endoscopy provide important prognostic information that 
guides subsequent management decisions (Fig. 51-1). Approximately 
20% of patients with bleeding ulcers have the highest-risk findings 
of active bleeding or a nonbleeding visible vessel; one-third of such 
patients have further bleeding that requires urgent surgery if they 
are treated conservatively. These patients benefit from endoscopic 
therapy such as bipolar electrocoagulation, heater probe, injection 
therapy (e.g., absolute alcohol, 1:10,000 epinephrine), clips, and/or 
topical hemostatic powder with reductions in bleeding, hospital stay, 
mortality, and costs. In contrast, patients with clean-based ulcers 
have rates of serious rebleeding approaching zero. If stable with no 
other reason for hospitalization, such patients may be discharged 
home after endoscopy.
Randomized controlled trials document that high-dose, proton 
pump inhibitor (PPI), given to reduce intragastric acid and thereby 
enhance clot stability, decreases further bleeding and mortality in 
patients with high-risk ulcers (active bleeding, nonbleeding visible ves­
sel, adherent clot) when given after endoscopic therapy. Meta-analysis of 
randomized trials indicates that outcomes are comparable with highdose PPIs given as a constant infusion or intermittently. Patients with 
lower-risk findings (flat pigmented spot, clean base) do not require 
endoscopic therapy and receive standard doses of oral PPI.
Approximately 10–50% of patients with bleeding ulcers rebleed 
within the next year if no preventive strategies are employed. Pre­
vention of recurrent bleeding focuses on the three main factors in 
ulcer pathogenesis, Helicobacter pylori, nonsteroidal anti-inflammatory 
drugs (NSAIDs), and acid. Eradication of H. pylori in patients with 
bleeding ulcers decreases rebleeding rates to <5%. If a bleeding ulcer 
develops in a patient taking NSAIDs, the NSAIDs should be discon­
tinued. If NSAIDs must be given, a cyclooxygenase-2 selective NSAID 
plus a PPI is recommended, based on results of a randomized trial. 
Patients with established cardiovascular disease who develop bleeding 
Endoscopic
Diagnosis
Ulcer
Erosions
Flat
pigmented
spot
Active
bleeding or
visible vessel
Endoscopic
Features
Adherent
clot
May consider
endoscopic
therapy
No
endoscopic
therapy
Endoscopic
Therapy
Endoscopic
therapy
Medical
Therapy
High-dose
PPI therapya
High-dose
PPI therapya
Once-daily
PPI therapy
Once-daily
PPI therapy
Clear liquids
for ~2 days
Clear liquids
for ~2 days
Clear liquids
for ~1 day
Regular
diet
Dietc
Hospital
Stayd
Hospitalize
3 days
Hospitalize
~1–2 days
Discharge after
endoscopy
Hospitalize
3 days
FIGURE 51-1  Suggested algorithm for patients with acute upper gastrointestinal bleeding (hematemesis, melena) based on endoscopic findings. aIntravenous bolus 
(80 mg) followed by infusion (8 mg/h) or by intermittent oral or intravenous doses (e.g., 40 mg 2–4 times per day) for 3 days. bIntravenous 50 μg bolus followed by 50 μg/h 
infusion for 2–5 days. cDiet after endoscopy, assuming no nausea or vomiting. dDuration after endoscopy assuming patient stable without further bleeding or concurrent 
medical conditions requiring hospitalization. PPI, proton pump inhibitor.

ulcers while taking low-dose aspirin for secondary prevention should 
not discontinue aspirin and, if aspirin is held, should restart aspi­
rin once hemostasis is confirmed. A randomized trial showed that 
immediate reinstitution of aspirin was associated with a lower 8-week 
mortality compared to not restarting aspirin (1% vs 13%; hazard ratio, 
0.2; 95% CI, 0.1–0.6). In contrast, aspirin should be discontinued in 
patients taking aspirin for primary prevention of cardiovascular events 
who develop UGIB. Patients with bleeding ulcers unrelated to H. pylori 
or NSAIDs should remain on PPI therapy indefinitely given a 42% inci­
dence of rebleeding at 7 years without protective therapy. Peptic ulcers 
are discussed in Chap. 335. 

Gastrointestinal Bleeding 
CHAPTER 51
MALLORY-WEISS TEARS  Mallory-Weiss tears account for ~2–10% of 
UGIB hospitalizations. The classic history is vomiting, retching, or 
coughing preceding hematemesis. Bleeding from these tears, which 
are usually on the gastric side of the gastroesophageal junction, stops 
spontaneously in ~80–90% of patients and recurs in only 0–10%. 
Endoscopic therapy is indicated for actively bleeding Mallory-Weiss 
tears. Mallory-Weiss tears are discussed in Chap. 334. 
ESOPHAGEAL VARICES  The proportion of UGIB hospitalizations 
due to varices varies widely from ~2–40%, depending on the popula­
tion. Patients with variceal hemorrhage have poorer outcomes than 
patients with other sources of UGIB. Esophageal varices are treated 
with endoscopic ligation and an IV vasoactive medication (octreotide, 
somatostatin, vapreotide, terlipressin) for 2–5 days. Combination of 
endoscopic and medical therapy is superior to either therapy alone in 
decreasing rebleeding. Over the long term, treatment with nonselective 
beta blockers plus endoscopic ligation is recommended because the 
combination is more effective than either alone in reduction of recur­
rent esophageal variceal bleeding. Transjugular intrahepatic portosys­
temic shunt (TIPS) is recommended in patients who have persistent 
or recurrent bleeding despite endoscopic and medical therapy. TIPS 
also is suggested for acute variceal bleeding in patients with advanced 
liver disease (Child-Pugh class B with score 8–9 and active bleeding at 
endoscopy, Child-Pugh class C with score 10–13), because randomized 
trials show significant decreases in rebleeding and mortality compared 
with standard endoscopic and medical therapy.
Esophageal Varices
Mallory-Weiss Tear
Clean
base
Active
bleeding
No active
bleeding
No
endoscopic
therapy
No
endoscopic
therapy
No
endoscopic
therapy
Endoscopic
ligation
Endoscopic
therapy
Antiemetic
if ongoing
nausea
Antiemetic
if ongoing
nausea
Vasoactive drug (e.g.,
octreotideb) + antibiotic
(e.g., ceftriaxone)
Once-daily
PPI therapy
Clear liquids
for ~2 days
Clear liquids
for ~1 day
Regular diet
Regular diet
Hospitalize
~3–5 days
Hospitalize
~1–2 days
Discharge after
endoscopy
Discharge after
endoscopy

Portal hypertension is also responsible for bleeding from gastric 
varices, varices in the small and large intestine, and portal hypertensive 
gastropathy and enterocolopathy. Bleeding gastric varices are treated 
with endoscopic injection of tissue adhesive (e.g., n-butyl cyanoacry­
late), TIPS, or retrograde transvenous obliteration.

EROSIVE DISEASE  Erosions are endoscopically visualized breaks that 
are confined to the mucosa and do not cause major bleeding because 
arteries and veins are not present in the mucosa. Erosions in the 
esophagus, stomach, or duodenum commonly cause mild UGIB, with 
erosive esophagitis (primarily due to gastroesophageal reflux disease), 
gastritis, and duodenitis accounting for up to perhaps ~30% of UGIB 
hospitalizations. The most important cause of gastric and duodenal 
erosions is NSAID use: up to ~50% of patients who chronically ingest 
NSAIDs may have gastric erosions. Other potential causes of gastric 
erosions include alcohol intake, H. pylori infection, and stress-related 
mucosal injury.
PART 2
Cardinal Manifestations and Presentation of Diseases
Stress-related gastric mucosal injury occurs only in extremely ill 
patients, such as those with serious trauma, major surgery, burns cov­
ering more than one-third of the body surface area, major intracranial 
disease, or severe medical illness. Severe bleeding should not develop 
unless ulceration occurs. The mortality rate in these patients is high 
because of their serious underlying illnesses.
The incidence of bleeding from stress-related gastric mucosal injury 
has decreased dramatically in recent years, most likely due to better care 
of critically ill patients. Meta-analysis of randomized trials indicates 
prophylaxis with PPIs reduces clinically important GIB more than no 
prophylaxis or H2-receptor antagonists without impacting mortality or 
risk of infection (e.g., pneumonia, Clostridioides difficile). A guideline 
suggested PPI prophylaxis in critically ill patients at high risk (≥4%) of 
bleeding, defined as mechanical ventilation without enteral nutrition, 
portal hypertension, cirrhosis, platelets <50 × 109/L, international nor­
malized ratio >1.5, or two of the following: mechanical ventilation with 
enteral nutrition, acute kidney injury, sepsis, or shock.
OTHER CAUSES  Less common causes of UGIB include neoplasms, 
vascular ectasias (including hereditary hemorrhagic telangiectasias and 
gastric antral vascular ectasia), Dieulafoy’s lesion (in which an aberrant 
vessel in the mucosa bleeds from a pinpoint mucosal defect), prolapse 
gastropathy (prolapse of proximal stomach into esophagus with retch­
ing), aortoenteric fistulas, and hemobilia or hemosuccus pancreaticus 
(bleeding from the bile duct or pancreatic duct).
Small-Intestinal Sources of Bleeding 
Patients without a source 
of GIB identified on upper endoscopy and colonoscopy were previ­
ously labeled as having obscure GIB. With the advent of improved 
diagnostic modalities, ~75% of GIB previously labeled obscure is now 
estimated to originate in the small intestine beyond the extent of a 
standard upper endoscopic exam. Small-intestinal GIB may account 
for ~5% of GIB cases. The most common causes in adults include 
vascular ectasias, neoplasm (e.g., gastrointestinal stromal tumor, car­
cinoid, adenocarcinoma, lymphoma, metastases), and NSAID-induced 
erosions and ulcers. Meckel’s diverticulum is the most common cause 
of significant small-intestinal GIB in children, decreasing in frequency 
as a cause of bleeding with age. Other less common causes of smallintestinal GIB include Crohn’s disease, infection, ischemia, vasculitis, 
small-bowel varices, diverticula, intussusception, Dieulafoy’s lesions, 
aortoenteric fistulas, and duplication cysts.
Small-intestinal vascular ectasias are treated initially with endo­
scopic therapy, based on observational studies suggesting short-term 
efficacy. However, rebleeding is common, with pooled rebleeding rates 
of ~45% over a mean follow-up of ~2 years in systematic reviews, lead­
ing guidelines to suggest medical therapy if further bleeding occurs 
after endoscopic therapy. The best available evidence supports use of 
thalidomide, with a multicenter double-blind randomized trial show­
ing marked reductions in bleeding episodes, transfusions, and hospi­
talizations. Monthly intramuscular injection of octreotide long-acting 
release also is suggested based on observational studies and a small 
open-label randomized trial. Other isolated lesions, such as tumors, 
generally require surgical resection.

Colonic Sources of Bleeding 
Hemorrhoids are probably the 
most common cause of lower GIB (LGIB); anal fissures also cause 
minor bleeding and pain. If these local anal processes, which rarely 
require hospitalization, are excluded, the most common cause of 
LGIB in adults is diverticulosis. Other causes include vascular ectasias 
(especially in the proximal colon of patients >70 years), neoplasms 
(primarily adenocarcinoma), colitis (ischemic, infectious, Crohn’s or 
ulcerative colitis, NSAID-induced colitis or ulcers), postpolypectomy 
bleeding, and radiation proctopathy. Rarer causes include solitary rec­
tal ulcer syndrome, varices (most commonly rectal), lymphoid nodular 
hyperplasia, vasculitis, trauma, and aortocolic fistulas. In children and 
adolescents, the most common colonic causes of significant GIB are 
inflammatory bowel disease and juvenile polyps.
Diverticular bleeding is abrupt in onset, usually painless, sometimes 
massive, and often from the right colon; chronic or occult bleeding 
is not characteristic. Case series from the United States and Europe 
suggest colonic diverticula stop bleeding spontaneously in ≥90% of 
patients, with rebleeding on long-term follow-up as low as ~15% over 
4–5 years. Rebleeding is substantially higher in reports from Asia. Case 
series suggest endoscopic therapy may decrease recurrent bleeding in 
the uncommon case when colonoscopy identifies the specific bleeding 
diverticulum. When diverticular bleeding is found at angiography, 
transcatheter arterial embolization by superselective technique pre­
vents further bleeding in most patients. Segmental surgical resection is 
recommended for refractory diverticular bleeding.
Bleeding from colonic vascular ectasias may be overt or occult; it 
tends to be chronic and only occasionally hemodynamically signifi­
cant. Endoscopic hemostatic therapy may be used in the treatment of 
vascular ectasias, as well as discrete bleeding ulcers and postpolyp­
ectomy bleeding. Transcatheter arterial embolization also may be 
attempted for persistent bleeding from vascular ectasias and other 
lesions, although rebleeding is higher in nondiverticular LGIB at ~45%. 
Surgical therapy is generally required for major persistent or recur­
rent bleeding from colonic sources that cannot be treated medically, 
endoscopically, or angiographically. Patients with Heyde’s syndrome 
(bleeding vascular ectasias and aortic stenosis) appear to benefit from 
aortic valve replacement.
APPROACH TO THE PATIENT
Gastrointestinal Bleeding
INITIAL ASSESSMENT
Measurement of the heart rate and blood pressure is the best way 
to initially assess a patient with GIB. Clinically significant bleeding 
leads to postural changes in heart rate or blood pressure, tachycar­
dia, and, finally, recumbent hypotension. In contrast, hemoglobin 
does not fall immediately with acute GIB, due to proportionate 
reductions in plasma and red cell volumes (“people bleed whole 
blood”). Thus, hemoglobin may be normal or only minimally 
decreased at initial presentation of a severe bleeding episode. As 
extravascular fluid enters the vascular space to restore volume, the 
hemoglobin falls, but this process may take up to 72 h. Transfusion 
is recommended when the hemoglobin drops below 7 g/dL, based 
on a large randomized trial showing this restrictive transfusion 
strategy decreases rebleeding and death in acute UGIB compared 
with a transfusion threshold of 9 g/dL. Patients with slow, chronic 
GIB may have very low hemoglobin values despite normal blood 
pressure and heart rate. With the development of iron-deficiency 
anemia, the mean corpuscular volume is low and red-blood-cell 
distribution width is increased. 
DIFFERENTIATION OF UGIB FROM LGIB
Hematemesis indicates an UGIB source. Melena indicates blood 
has been present in the gastrointestinal (GI) tract for ≥14 h and as 
long as 3–5 days. The more proximal the bleeding site, the more 
likely melena will occur. Hematochezia usually represents a lower 
GI source of bleeding, although an upper GI lesion may bleed so 
briskly that blood transits the bowel before melena develops. When

hematochezia is the presenting symptom of UGIB, it is associated 
with hemodynamic instability and dropping hemoglobin. Bleeding 
lesions of the small bowel may present as melena or hematochezia. 
Other clues to UGIB include hyperactive bowel sounds and elevated 
blood urea nitrogen (due to volume depletion and blood proteins 
absorbed in the small intestine).
A nonbloody nasogastric aspirate may be seen in ~15% of 
patients with UGIB who present with clinically serious hemato­
chezia. A bile-stained appearance does not exclude UGIB because 
reports of bile in the aspirate are incorrect in ~50% of cases. Testing 
of aspirates that are not grossly bloody for occult blood is not useful. 
EVALUATION AND MANAGEMENT OF UGIB (FIG. 51-1)
Initial Risk Assessment  Baseline characteristics predictive of rebleed­
ing and death include hemodynamic compromise (tachycardia or 
hypotension), increasing age, and comorbidities. Risk assessment tools 
may be used to identify patients with very low risk. Discharge from the 
emergency room with outpatient management is suggested for patients 
with a Glasgow-Blatchford score (possible range 0–23, Table 51-1) of 
0–1 because only ~1% of patients who require transfusion, require 
hemostatic intervention, or die have a score of 0–1. 
Pre-Endoscopic Medications  PPI infusion may be considered 
at presentation; it modestly decreases need for endoscopic therapy 
due to a reduction in high-risk ulcer stigmata (e.g., active bleeding) 
but does not improve clinical outcomes such as further bleeding or 
death. The promotility agent erythromycin, 250 mg intravenously 
~30–90 min before endoscopy, is suggested to improve visualization 
at endoscopy, thereby reducing the need for repeat endoscopy and 
hospital stay. Cirrhotic patients presenting with UGIB should be 
given an antibiotic (e.g., ceftriaxone) and IV vasoactive medication 
(e.g., octreotide) upon presentation. Antibiotics decrease bacterial 
infections, rebleeding, and mortality, and vasoactive medications 
improve control of bleeding in the 12 h after presentation. 
Endoscopy  Upper endoscopy should be performed within 24 h 
in most patients hospitalized with UGIB whether they have clinical 
No Hemodynamic Instability
Hemodynamic Instability
Colonoscopya
Site identified,
bleeding persists
Site not
identified
Site identified,
bleeding stops
Angiography
Bleeding persists
Surgery
FIGURE 51-2  Suggested algorithm for patients with acute lower gastrointestinal (GI) bleeding (hematochezia). aFor patients <35 years old with minor bleeding (e.g., 
blood on toilet paper), normal blood pressure and heart rate, normal hemoglobin and iron panel, and no family history of colorectal cancer, some would suggest flexible 
sigmoidoscopy is adequate.

TABLE 51-1  Glasgow-Blatchford Score
RISK FACTORS AT ADMISSION
SCORE
Blood urea nitrogen (mg/dL)
  18.2 to <22.4

  22.4 to <28.0

  28.0 to <70.0

  ≥70.0

Hemoglobin (g/dL)
  12.0 to <13.0 (men); 10.0 to <12.0 (women)

Gastrointestinal Bleeding 
CHAPTER 51
  10.0 to <12.0 (men)

  <10.0

Systolic blood pressure (mmHg)
  100–109

  90–99

  <90

Heart rate (beats per minute)
  ≥100

Melena

Syncope

Hepatic disease

Cardiac failure

features predicting low or high risk of further bleeding and death. 
Even in high-risk patients, more urgent endoscopy (performed 
within 6–12 h of gastroenterology consultation) does not improve 
clinical outcomes, and some observational studies suggest increased 
mortality with endoscopy within 6–12 h in high-risk patients. 
Early endoscopy in low-risk patients (e.g., hemodynamically stable 
without severe comorbidities) identifies low-risk findings (e.g., 
clean-based ulcers, erosions, nonbleeding Mallory-Weiss tears) that 
allow discharge in ≥40% of patients, thereby reducing hospital stay 
Upper endoscopy
No upper GI source
Able to prep
Too unstable to prep
No
Extravasation
CT angiography
Able to prep
Extravasation
Angiography
Bleeding persists
Instability persists
Workup for small
intestinal/obscure
bleeding site
Surgery (with intraoperative
endoscopy if site has not
been identified)