Digestive Diseases and Sciences (2018) 63:1286–1293

https://doi.org/10.1007/s10620-017-4882-6

ORIGINAL ARTICLE

Changing Epidemiology of Upper Gastrointestinal Hemorrhage

in the Last Decade: A Nationwide Analysis
Brandon A. Wuerth1 · Don C. Rockey1

Received: 18 November 2016 / Accepted: 8 December 2017 / Published online: 27 December 2017
© Springer Science+Business Media, LLC, part of Springer Nature 2017

Abstract
Background  Upper gastrointestinal hemorrhage (UGIH) is common and carries substantial mortality requiring frequent
hospitalizations.
Aim  To investigate trends in etiology and outcome of UGIH in hospitalized patients in the USA.
Methods  Retrospective, observational cohort study of the Nationwide Inpatient Sample from 2002 to 2012 was carried out.
UGIH was identified in hospitalizations with a principle ICD-9-CM diagnosis of UGIH or secondary diagnosis of UGIH
with a principal diagnosis of hematemesis, blood in stool, or gastrointestinal bleeding. Age 18 years or older was required
for inclusion, and elective admissions and transferred patients were excluded.
Results  The hospitalization rate of UGIH in the USA decreased by 21% from 2002 to 2012, from 81 to 67 cases per 100,000
population (p < 0.01). The greatest declines occurred for gastritis and PUD, which decreased by 55 and 30%, respectively
(p < 0.01). There were increases in neoplasm, Dieulafoy lesions, angiodysplasia, and esophagitis, which increased by 50, 33,
32 and 20%, respectively (p < 0.01). The all-cause inpatient mortality rate of UGIH decreased 28% from 2.6 per 100 cases
in 2002 to 1.9 in 2012 (p < 0.01). The greatest decline occurred for esophagitis, Mallory–Weiss tear, and neoplasm, which
decreased by 39% (p < 0.01), 36% (p = 0.02), and 36% (p < 0.01), respectively. The rate of hospitalization for bleeding
caused by esophageal varices remained constant and low (approximately 2%) throughout the study period; the mortality for
esophageal varices also remained constant at 6–7%.
Conclusions  The epidemiology of UGIH hemorrhage appears to be shifting, with a decline in PUD and gastritis; an increase
in hospitalization rate for neoplasm, Dieulafoy lesions, angiodysplasia, and esophagitis; and a reduction in overall mortal-
ity. The decreasing hospitalization rate and mortality rate of UGIH suggest population trends in use of treatments for PUD,
improved hemostatic techniques, and overall care.

Keywords  Bleeding · Mortality · Outcome · Etiology

Introduction

Upper gastrointestinal hemorrhage (UGIH) in the USA

Data reported in part as oral presentation at Digestive Diseases
Week, May 23rd, 2016.
Electronic supplementary material  The online version of this
article (http​s://doi.org/10.1007​/s106​20-017-4882​-6) contains
supplementary material, which is available to authorized users.
* Don C. Rockey
rockey@musc.edu
1
Division of Gastroenterology, Department of Medicine,
Medical University of South Carolina, 96 Jonathan Lucas
Street, Suite 803, MSC 623, Charleston, SC 29425, USA
requires frequent hospitalizations and carries a substantial
morbidity and mortality with over $1 billion in direct medi-
cal costs annually [1, 2]. UGIH has a variety of etiologies
including esophageal varices, peptic ulcer disease (PUD),
esophagitis, and gastritis, among others. The different causes
of UGIH each carry with it different epidemiologic features
and mortality rates. Historically, PUD has been the most
common cause of UGIH (31–67%), while variceal bleed-
ing has been shown to have the highest mortality (11–50%)
[3–5]. Currently, evidence suggests that both the epidemiol-
ogy of the etiology of UGIH and the mortality associated

13
Vol:.(1234567890)
Digestive Diseases and Sciences (2018) 63:1286–1293 1287
with it are dynamic and have changed over the last 30 years
[2, 6, 7].
Innovations in endoscopic hemostatic techniques have
evolved over the past 30 years and are thought to have par-
tially contributed to the decrease in mortality of UGIH [8].
Additionally, the widespread use of proton pump inhibitors
for PUD has been proposed to have not only affected the
epidemiology of etiologies of UGIH, but also its outcome
[9, 10].
The objective of this study was to investigate epidemio-
logic trends in clinical features, causes of, and outcomes of
patients admitted to the hospital with UGIH using a large
nationwide database.
Methods
Study Design
This was a retrospective cohort study examining hospitaliza-
tions with UGIH using discharge data from the Nationwide
Inpatient Sample (NIS) database [11]. Healthcare Cost and
Utilization Project (HCUP) databases are based on statewide
data collected by individual data organizations that provide
it to Agency for Healthcare Research and Quality (AHRQ).
The NIS database contains an approximately 20% stratified
probability sample of all US hospitals. Based on the 20%
sample, AHRQ calculated national estimates by weighting
each discharge based upon several factors relative to that of
all US hospitals. Estimations of total US hospitalizations
were made by using these calculated discharge weights. The
NIS database includes hospital data, patient demographic
data, and discharge data including up to 25 discharge Inter-
national Classification of Disease, 9th revision, clinical
modification (ICD-9-CM) codes.
The study was deemed “Not Human Subjects Research”
by the institutional review board and was therefore exempted
from review (Protocol #5267). The purchaser and beholder
(BAW) of the database completed the required HCUP Data
Use Agreement Training.
Case Selection
We included patients who were 18 years or older and dis-
charged with an ICD-9 principal diagnosis of UGIH (ICD-9
codes used are shown in Supplemental Table 1) or secondary
diagnosis of UGIH with an associated principle diagnosis
of hematemesis (578.0), blood in stool (578.1), or hemor-
rhage of gastrointestinal tract (unspecified) (578.9) as out-
lined in Fig. 1. Elective admissions and patients that were
transferred to another short-term hospital were excluded. We
additionally performed an analysis in which an EGD was
additionally performed (this substantively changed only a
few of the reported results, see Results section).
The ICD-9 code for Dieulafoy lesion was not available
until late 2002; therefore, data for Dieulafoy lesion in 2003
was also used for 2002. One of the ICD-9 codes used for
esophagitis (530.21) was not available until late 2003; there-
fore, data for this code in 2004 was also used for 2002 and
2003. Gastric varices were not included in the study as there
was no specific ICD-9 code for such. The control group of
non-UGIH cohort consisted of those hospitalizations with-
out any diagnosis of UGIH, hematemesis, blood in stool, or
hemorrhage of gastrointestinal tract.
Analysis
For each discharge, we recorded all listed diagnoses, age,
sex, race, hospital teaching status, payer status, and whether
the patient died from any cause during the hospitalization.
Population estimates were obtained from the U.S. Census
Bureau based on the annual population estimate from July
1st of each year [12].
Principal outcomes were temporal trends in the hospi-
talization rate and all-cause inpatient mortality rate of each
etiology in hospitalized patients. Hospitalization rate was
calculated using the number of hospitalizations from the
NIS database as the numerator and the population estimates
from the U.S. Census Bureau as the denominator. Similarly,
we calculated the all-cause mortality rate using the number
of deaths in the numerator and the number of cases in the
denominator.
Data were extracted from the NIS using SAS v9.4 (SAS
Institute Inc., Cary, NC, USA). Trends over the study period
were calculated as percent changes using linear regression
and was performed using GraphPad Prism v6.04 (GraphPad
Software Inc., La Jolla, CA). Trends between UGIH and
non-UGIH cases were calculated using XY linear regres-
sion tables with discharges as the dependent variable (Y
columns) and year as the independent variable (X column).
P values were calculated to determine whether the slopes of
UGIH and non-UGIH cases were significantly different. P
values less than 0.05 were considered significant. All other
descriptive statistics were completed using Microsoft Excel
2016 (Microsoft Corporation, Redmond, WA, USA).
Results
From 2002 to 2012, there were 2,432,088 cases of UGIH
and 235,516,630 cases in the control group of non-UGIH
(Table 1). Peptic ulcer disease comprised 47% of all cases
followed by gastritis and esophagitis, 18 and 15%, respec-
tively. Compared to control, UGIH tended to occur in an
older and male population with 44% of all cases being
13

1288 Digestive Diseases and Sciences (2018) 63:1286–1293

Table 1  Study demographics UGIH (n = 2,432,088) Non-UGIH

(n = 235,516,630)

Etiology
 Peptic ulcer 47.1% –
 Gastritis 18.1% –
 Esophagitis 15.2% –
 Angiodysplasia 6.2% –
 Mallory–Weiss 6.9% –
 Neoplasm 3.7% –
 Esophageal varices 1.8% –
 Dieulafoy lesions 1.5% –
Age
 18–44 12% 29%
 45–64 30% 28%
 65–84 44% 32%
 ≥ 85 14% 11%
Gender
 Male 55% 41%
 Female 45% 59%
Race
 White 71% 68%
 Black 13% 15%
 Hispanic 9% 11%
 Asian 3% 2%
 Native am 1% 1%
 Other 2% 3%
Payer
 Medicare 60% 48%
 Medicaid 8% 15%
 Private 22% 28%
 Self pay 6% 6%
 No charge 1% 1%
 Other 3% 3%
Teaching status
 Rural 13% 13%
 Urban non-teaching 47% 43%
 Urban teaching 40% 44%
Elixhauser Comorbidity Index (mean) 2.9 2.1
Length of stay [mean (days)] 4.7 4.9

between 65 and 84 years old and 55% being male. Upper
gastrointestinal hemorrhage cases also had higher mean
Elixhauser Comorbidity Index (ECI) scores while having
a similar mean length of stay (LOS) of 5 days compared to
non-UGIH cases.
The overall hospitalization rate of UGIH in the USA
decreased from 81 to 67 cases per 100,000 population from
2002 to 2012 (Table 2). Using linear regression, this was a
21% decrease (p < 0.01) while the hospitalization rate of
non-UGIH cases decreased 1% (p = 0.51). Figure 2a shows
the hospitalization rate from 2002 to 2012 normalized to
2002 to be able to show percent changes from year to year.
As shown, the declining trend in UGIH outpaced that of
non-UGIH cases (p < 0.01). In order to enhance the speci-
ficity of diagnosis, we performed an additional analysis in
which EGD was also required for inclusion; with this analy-
sis, the decrease in the overall hospitalization rate of UGIH
was 18% (i.e., compared to 21% when EGD was not required
as above) and the decrease in non-UGIH hospitalization was
6% (p < 0.01).
The all-cause case fatality rate of UGIH decreased from
2.6 to 1.9 per 100 cases (Table 3). Using linear regression,

13
Digestive Diseases and Sciences (2018) 63:1286–1293 1289

Fig. 1  Inclusion and exclusion criteria for UGIH and non-UGIH (control) groups. EGD was required for inclusion, and secondary diagnoses of
UGIH required a principal diagnosis of gastrointestinal hemorrhage to be included

Table 2  Hospitalization rate (cases per 100,000 population) of UGIH

by etiology
Hospitalization rate (cases per 100,000 popula-
tion)
2002 2012 % Change p value

PUD 41 30 − 30 < 0.01

Gastritis 17 10 − 55 < 0.01
Esophagitis 10.6 12.2 20 < 0.01
Angiodysplasia 4 5 32 < 0.01
Mallory–Weiss 4.9 5.1 1 0.70
Neoplasm 2.2 3.2 50 < 0.01
E. varices 1.5 1.4 − 5 0.26
Dieulafoy 1.0 1.2 33 < 0.01
UGIH (total) 81 67 − 21 < 0.01
Non-UGIH 7108 6907 − 1 0.51

this was a 28% decrease which was more than the decrease
in the mortality rate of non-UGIH cases at only 23%
(p < 0.01). Figure 2b shows the hospitalization rate from
2002 to 2012, again normalized to 2002 to be able to show
percent changes from year to year. The declining trend in
UGIH mortality outpaced that of non-UGIH cases, but
the difference was not statistically significantly different
(p = 0.19). As above, when an EGD was also required for Fig. 2  Hospitalization rates and mortality from 2002 to 2012. In a is
inclusion, the overall decrease in the overall mortality rate shown normalized hospitalization rate (cases per 100,000 population)
of UGIH compared to non-UGIH cases. The dotted lines indicate
of UGIH was the same (2.2–1.5 per 100 cases). For spe-
95% CIs. In b is shown normalized mortality rate (cases per 100,000
cific diagnoses, the magnitude of change over time were the population) of UGIH compared to non-UGIH cases. The dotted lines
same, with the exception that the decrease in mortality of indicate 95% CIs

13

1290 Digestive Diseases and Sciences (2018) 63:1286–1293

Table 3  All-cause mortality rate (deaths per 100 cases) of UGIH by per 100,000 population) as the most common causes of
etiology UGIH in 2002. However, from 2002 to 2012, the hospitali-
Mortality (deaths per 100 cases) zation rate of gastritis decreased 55% and the hospitalization
rate of esophagitis increased 20%; thus, by 2012 esophagitis
2002 2012 % Change p value
was the second most common cause of UGIH.
E. varices 7.3 6.1 0.2 0.98 From 2002 to 2012, the hospitalization rate of PUD
Neoplasm 6.9 5.1 − 36 < 0.05 decreased 30% from 41 to 30 cases per 100,000 popula-
Dieulafoy 3.8 2.9 − 26 < 0.05 tion (Fig. 3a). As for trends in other etiologies, the hos-
PUD 2.8 2.0 − 32 < 0.01 pitalization rate of esophageal variceal hemorrhage and
Mallory–Weiss 2.0 1.3 − 36 < 0.01 Mallory–Weiss tear changed minimally (− 5 and + 1%,
Esophagitis 2.0 1.4 − 39 < 0.01 respectively, Fig. 3b). Dieulafoy lesions, angiodysplasia, and
Gastritis 1.6 1.3 − 21 < 0.05 neoplasm increased 33, 32, and 50%, respectively. Changes
Angiodysplasia 1.5 1.0 − 26 < 0.01 in the hospitalization rates of Mallory–Weiss tear and esoph-
UGIH (total) 2.6 1.9 − 28 < 0.01 ageal varices were not significant over the study period.
Non-UGIH 3.2 2.5 − 23 < 0.01 The mortality rate in all of the different etiologic cat-
egories of UGIH declined from 2012 compared to 2002,
but mortality differences for esophageal variceal hemor-
gastritis was 1.4–1.3 per 100 cases and was not statistically rhage over time were not significantly different. The largest
significantly different (p = 0.13). improvements in mortality occurred for the diagnoses of
Regarding etiology, PUD was followed by gastritis (17 esophagitis, neoplasm, and Mallory–Weiss tear (decreases
cases per 100,000 population) and then esophagitis (11 cases of 39, 36, and 36%, respectively, Fig. 4).

Fig. 3  Hospitalization rates for

specific UGI bleeding disorders
between 2002 and 2012. In a is
shown the hospitalization rate
(cases per 100,000 population)
of peptic ulcer disease, gastritis
and esophagitis. In b is shown
the hospitalization rate (cases
per 100,000 population) for
angiodysplasia, Mallory–Weiss
tears, neoplasm, esophageal
variceal hemorrhage, and Dieu-
lafoy lesions

13
Digestive Diseases and Sciences (2018) 63:1286–1293 1291
Fig. 4  Mortality (from 2002 to
2012) in different UGI disease
categories. The mortality rate
(deaths per 100 cases) for each
disease for each year
Discussion
Using data from a large national database, we have found
that from 2002 to 2012 the hospitalization rate of UGIH
in the USA decreased. This was driven primarily by large
decreases in the hospitalization rate of PUD and gastritis,
as the hospitalization rate of esophagitis, Dieulafoy lesions,
angiodysplasia, and neoplasm increased. As of 2012, PUD
remained the most common cause of UGIH; however,
esophagitis surpassed gastritis as the second most common
cause. The mortality rate of UGIH decreased significantly
for all diseases from 2002 to 2012 except for esophageal
varices, with the greatest decline in esophagitis, Mal-
lory–Weiss tear, and neoplasm.
Our results are consistent with but also differ from other
studies examining UGIH. One such study using the Premier
database from 2001 to 2009 found a 23% decrease in hos-
pitalization rate from 78 to 61 cases per 100,000 popula-
tion and a 17% decrease in the mortality rate (from 2.95 to
2.45%) [2]. Our study also found a decrease in the hospitali-
zation rate and mortality rate, and while our observed mor-
tality rate was lower than observed in this previous study, the
differences do not appear to be substantial. However, there
were several notable differences among the studies that build
upon the current literature of UGIH.
First, we included esophagitis as a diagnosis in our cohort
while the previous database study did not. We demonstrate
that esophagitis not only makes up 15% of the overall cases
of UGIH, but has also become increasingly more common.
Secondly, our study also included cases with a principal
diagnosis of undefined UGIH if there was a secondary
diagnosis of defined UGIH while the previous study only
included principal diagnoses.
A previous study utilizing the NIS database found that
UGIH mortality decreased from 4.7% in 1989 to 2.1% in
2009 while the hospitalization rate decreased from 83 to
79 cases per 100,000 population [13], consistent with the
findings from our study. However, data were only analyzed
at 5-year intervals and was over a different time period
than the current study. Further, this study only stratified
bleeding into 2 categories—variceal UGIH and non-
variceal UGIH. Stratifying by each individual etiology
helps elicit apparent differences in epidemiology.
In terms of the individual causes of UGIH, our data dif-
fer from other recent studies. For example, several studies
conducted at large safety net hospitals demonstrated a very
high hospitalization rate of variceal hemorrhage [14–16].
In contrast, we report that in each 2002 and 2012, the
overall hospitalization rate of esophageal variceal hem-
orrhage was only roughly 2% of UGIH cases [14]. This
may represent the true prevalence of esophageal variceal
hemorrhage at a population level, since many reports about
esophageal variceal hemorrhage come from in large urban
medical centers and not community hospitals. However,
this may also be an underrepresentation of the true inci-
dence of esophageal variceal hemorrhage for several rea-
sons. First, we were not able to include gastric variceal
hemorrhage as there was no specific ICD-9 diagnosis for
this disease—and this may have led to a reduction in the
frequency of variceal hemorrhage reported here. It is also
possible that variceal bleeding may have been coded as
hematemesis without a specific diagnosis in cases where
varices were not actively bleeding but had high-risk signs
such as large varices, red whale sign, or white nipple sign.
Finally, we included all patients with GI bleeding without
the requirement for EGD. This may have led to an increase
in bleeding (presumably) caused by trivial upper gastroin-
testinal tract lesions.
While PUD has been reported to be relatively common
in previous studies, it was even more prevalent in the cur-
rent analysis. We found that nationwide PUD (47% overall
hospitalization rate) was by far the most prevalent cause of
UGIH. Gastritis (18%) and esophagitis (14%) were preva-
lent in all of the studies. The data presented with this large
national sample are thus important because they highlight
the possibility of selection bias when using smaller sample
13

1292
sizes that may explain differences in the causes of UGIH and
thus may not be representative of the US population.
It is important to note that our study had similar mortal-
ity rates to other nationwide studies when all patients with
UGIH are taken into account [2, 13]. However, previous
studies examining mortality in patient cohorts with specific
diagnoses (i.e., varies, PUD or esophageal varices) have gen-
erally reported much higher mortality rates ([14–17]). We
speculate that this likely represents selection bias; most of
these previous studies investigating specific causes of UGIH
were conducted in large urban centers (Los Angeles, Dal-
las, etc), where the underlying severity of illness is likely
to be high, while the NIS is representative of the entire US
population. This further raises interesting questions about
the epidemiology of UGIH in different types of hospitals
(i.e., large urban centered hospitals and smaller and rural
hospitals).
The decrease in UGIH due to PUD has primarily been
attributed to decreases in H. pylori prevalence and increased
use of acid suppression medications such as proton pump
inhibitors (PPI) [18]. The first oral PPI was introduced in
1989, while the first intravenous PPI was introduced in 2001.
NSAIDs are another important risk factor for PUD, and the
introduction of COX-2 inhibitors may have also influenced
the hospitalization rate of PUD which were first introduced
in late 2000.
Improved diagnostic and hemostatic techniques may also
contribute to the changes in epidemiology of UGIH over
the study period. A similar study using the NIS from 1989
to 2009 found that EGD was performed in 85% of UGIH
cases in 2009 while only 70% in 1989 [13]. In addition,
endoscopic therapy was utilized in 27% of UGIH cases in
2009 compared to only 10% in 1989. The increases seen in
EGDs performed and endoscopic therapy used, however,
could either be from improved management or increased
prevalence of UGIH cases. The study did, however, find that
EGDs were performed earlier in hospitalizations in 2009
than 1989 which could explain the overall improved mor-
tality. Also of note, the results of the study were consistent
with our current study; however, they did not break down
mortality by etiology other than variceal versus non-variceal
hemorrhage.
We recognize limitations of this study. First, we were una-
ble to validate ICD-9 codes, which was our case identifier,
through individual chart review. This allowed for potential
misclassification into our results. This could have caused an
error of omitting those who actually had UGIH (false nega-
tive) or by counting those who actually didn’t (false posi-
tive). For example, variceal bleeding may have been coded
as hematemesis without a specific diagnosis in cases where
varices were not actively bleeding but had high-risk signs
such as large varices, red whale sign, or white nipple sign.
However, a number of previous studies have shown good
13
Digestive Diseases and Sciences (2018) 63:1286–1293
predictive values of these codes [19–22]. It should be noted
that mortality rates were based on all causes, so it was not
possible to determine whether or not a direct complication
of UGIH was the cause of death.
In summary, the hospitalization rate and all-cause mortal-
ity rate of upper gastrointestinal hemorrhage appear to be
steadily decreasing in the USA. Esophageal variceal hemor-
rhage continues to be associated with the highest mortality
rate; however, there were no significant changes in either
hospitalization or mortality rate from 2002 to 2012. PUD
and gastritis as causes of UGIH continue to decline over-
all. The hospitalization rate of several diseases, including
esophagitis, Dieulafoy lesions, angiodysplasia, and neo-
plasm, appear to be increasing. Recognition of these features
should assist practitioners in the management of patients
with UGIH.
Author’s contribution  Brandon A. Wuerth were involved in the study
of the concept and design; acquisition of data; analysis and interpreta-
tion of data; drafting of the manuscript; critical revision of the manu-
script for important intellectual content. Don Rockey helped in the
study of the concept and design; analysis and interpretation of data;
drafting of the manuscript; critical revision of the manuscript for
important intellectual content, supervisory efforts.
Compliance with ethical standards 
Conflict of interest  The authors have no conflict of interest or funding
related to this study.
References
1. Esrailian E, Gralnek IM. Nonvariceal upper gastrointestinal bleed-
ing: epidemiology and diagnosis. Gastroenterol Clin North Am.
2005;34:589–605.
2. Laine L, Yang H, Chang SC, et al. Trends for incidence of hospi-
talization and death due to GI complications in the United States
from 2001 to 2009. Am J Gastroenterol. 2012;107:1190–1195.
(quiz 1196).
3. van Leerdam ME. Epidemiology of acute upper gastrointestinal
bleeding. Best Pract Res Clin Gastroenterol. 2008;22:209–224.
4. Rockall TA, Logan RF, Devlin HB, et al. Variation in outcome
after acute upper gastrointestinal haemorrhage. The national
audit of acute upper gastrointestinal haemorrhage. Lancet.
1995;346:346–350.
5. Rotondano G. Epidemiology and diagnosis of acute nonvariceal
upper gastrointestinal bleeding. Gastroenterol Clin North Am.
2014;43:643–663.
6. Longstreth GF. Epidemiology of hospitalization for acute upper
gastrointestinal hemorrhage: a population-based study. Am J Gas-
troenterol. 1995;90:206–210.
7. van Leerdam ME, Vreeburg EM, Rauws EA, et al. Acute upper GI
bleeding: did anything change? Time trend analysis of incidence
and outcome of acute upper GI bleeding between 1993/1994 and
2000. Am J Gastroenterol. 2003;98:1494–1499.
8. Cook DJ, Guyatt GH, Salena BJ, et al. Endoscopic therapy for
acute nonvariceal upper gastrointestinal hemorrhage: a meta-
analysis. Gastroenterology. 1992;102:139–148.
Digestive Diseases and Sciences (2018) 63:1286–1293 1293
9. Bardou M, Toubouti Y, Benhaberou-Brun D, et al. Meta-analysis:
proton-pump inhibition in high-risk patients with acute peptic
ulcer bleeding. Aliment Pharmacol Ther. 2005;21:677–686.
10. Leontiadis GI, Sharma VK, Howden CW. Proton pump inhibitor
treatment for acute peptic ulcer bleeding. Cochrane Database Syst
Rev. 2006;3:CD002094.
11. HCUP Databases. Healthcare cost and utilization project (HCUP).
Nationwide inpatient sample (NIS) aLdaahwh-uag. HCUP Data-
bases. Healthcare Cost and Utilization Project (HCUP). Nation-
wide Inpatient Sample (NIS), 2002 and 2012. Limited database
available at: http​://www.hcup​-us.ahrq​.gov.
12. United States Census Bureau. Annual population estimates avail-
able at: http​s://fact​find​er.cens​us.gov.
13. Abougergi MS, Travis AC, Saltzman JR. The in-hospital mortal-
ity rate for upper GI hemorrhage has decreased over 2 decades
in the United States: a nationwide analysis. Gastrointest Endosc.
2015;81(882–8):e1.
14. Kim JJ, Sheibani S, Park S, et al. Causes of bleeding and outcomes
in patients hospitalized with upper gastrointestinal bleeding. J
Clin Gastroenterol. 2014;48:113–118.
15. Balderas V, Bhore R, Lara LF, et al. The hematocrit level in upper
gastrointestinal hemorrhage: safety of endoscopy and outcomes.
Am J Med. 2011;124:970–976.
16. Guntipalli P, Chason R, Elliott A, et al. Upper gastrointestinal
bleeding caused by severe esophagitis: a unique clinical syn-
drome. Dig Dis Sci. 2014;59:2997–3003.
17. Lyles T, Elliott A, Rockey DC. A risk scoring system to predict in-
hospital mortality in patients with cirrhosis presenting with upper
gastrointestinal bleeding. J Clin Gastroenterol. 2014;48:712–720.
18. Perez-Aisa MA, Del Pino D, Siles M, et al. Clinical trends in ulcer
diagnosis in a population with high prevalence of Helicobacter
pylori infection. Aliment Pharmacol Ther. 2005;21:65–72.
19. Raiford DS, Perez Gutthann S, Garcia Rodriguez LA. Positive
predictive value of ICD-9 codes in the identification of cases of
complicated peptic ulcer disease in the Saskatchewan hospital
automated database. Epidemiology. 1996;7:101–104.
20. Cattaruzzi C, Troncon MG, Agostinis L, et al. Positive predic-
tive value of ICD-9th codes for upper gastrointestinal bleeding
and perforation in the Sistema Informativo Sanitario Regionale
database. J Clin Epidemiol. 1999;52:499–502.
21. Pisa F, Castellsague J, Drigo D, et al. Accuracy of international
classification of diseases, 9th revision, clinical modification codes
for upper gastrointestinal complications varied by position and
age: a validation study in a cohort of nonsteroidal anti-inflamma-
tory drugs users in Friuli Venezia Giulia, Italy. Pharmacoepide-
miol Drug Saf. 2013;22:1195–1204.
22. Cooper GS, Chak A, Lloyd LE, et al. The accuracy of diagnosis
and procedural codes for patients with upper GI hemorrhage. Gas-
trointest Endosc. 2000;51:423–426.
13