THE AMERICAN JOURNAL OF GASTROENTEROLOGY
1, 2000
© 2000 by Am. Coll. of Gastroenterology
Published by Elsevier Science Inc.
The Description of Outcomes in Medicare Patients
Hospitalized With Peptic Ulcer Disease
Wen Xu, Ph.D., A.S.A., Hugh M. Hood, M.D., F.A.C.P., and Patricia A. Burgess, R.N., B.S.N.
Alabama Quality Assurance Foundation, Birmingham, Alabama
OBJECTIVE: The objective of this study was to describe
outcomes of care for Medicare patients hospitalized with
peptic ulcer disease from 1992 through 1997 and to identify
factors related to cost, length of stay, and readmission rates.
METHODS: General descriptive statistics were obtained from
Medicare inpatient claims data by year, endoscopy group-
ing, diagnosis related group code, and principal diagnosis
code. From abstracted clinical data, associations were de-
rived for length of stay, readmission rates, and the following
processes of care: screening or treatment for Helicobacter
pylori; screening for nonsteroidal antiinflammatory drug
(NSAID) use; and the performance of endoscopy. The
Acute Physiology and Chronic Health Evaluation method
was used to estimate patient health status for the study.
RESULTS: During the 6-yr study, there were 878,212 claims,
which constituted 1.3% of the total Medicare claims. The
total Medicare payment for peptic ulcer claims was esti-
mated at $4.8 billion. The inpatient mortality rate was 4.5%.
Readmission rates remained relatively constant during the
study period but decreased significantly when NSAID
screening was documented during the hospitalization. Ad-
mission rates, length of stay, and mortality declined pro-
gressively during the study period. A reduction in length of
stay of approximately 1 day was observed when screening
or treatment for H. pylori, screening for NSAID use, or the
performance of endoscopy was documented.
CONCLUSIONS: Peptic ulcer disease has an important impact
on the Medicare population with respect to cost, recurrence,
and mortality. Adherence to selected processes of care is
associated with shorter length of stay and lower readmission
rates. (Am J Gastroenterol 2000;95:264 –270. © 2000 by
Am. Coll. of Gastroenterology)
INTRODUCTION
Until 1984, peptic ulcer disease was considered an idio-
pathic illness. Nearly 13 yr have elapsed since Marshall et
al. satisfied Koch’s postulates and established a so-called
“curved bacillus” as an etiological agent for acute gastritis
(1, 2). With the explosion of research in this area, the curved
bacillus has been classified as Helicobacter pylori (H. py-
lori) and is recognized as the most common chronic bacte-
rial infection in humans (3, 4). The organism has been
Vol. 95, No.
ISSN 0002-9270/00/$20.00
PII S0002-9270(99)00752-2
associated with duodenal ulcer in ⬎90% of cases and with
gastric ulcer in 60 – 80% of cases (5).
Several position papers have been published establishing
recommendations for diagnosis and treatment. In 1994, the
National Institute of Health Consensus Development Con-
ference suggested screening patients with peptic ulcer dis-
ease (PUD) for H. pylori and eradicating the organism if
present (6). In 1996, the American College of Gastroenter-
ology had a similar recommendation for treatment of H.
pylori, in addition to screening for nonsteroidal antiinflam-
matory drug (NSAID) use and eliminating or reducing the
dose if possible (7). The Digestive Health Initiative (DHI)
Update Conference in February 1997 further solidified the
existing state of knowledge of pathogenesis, epidemiology,
diagnosis, disease associations, and treatment of H. pylori.
Priorities for research needs were also formulated at this
international conference (4).
Despite these recommendations, recurrent PUD remains
common. More than 500,000 new cases are diagnosed each
year in the United States and four million people experience
recurrence (8). The cost to Medicare for hospitalizations
directly related to PUD is significant, with related claims for
approximately $800 million being processed each year (9).
Although PUD affects all age groups, it heavily affects the
Medicare population, as infection with H. pylori increases
with age (10). The use of NSAIDs also increases with age
(11). Without assessment of H. pylori or NSAIDs as the
etiology, PUD results in detrimental health outcomes such
as recurrent ulcers, gastrointestinal (GI) bleeding, repeated
hospitalizations, unnecessary surgical procedures, increased
drug expense, chronic gastritis, gastric cancer, and increased
mortality (12, 13).
In May 1996, the Health Care Financing Administration
(HCFA) initiated a multistate project to improve the quality
of care for Medicare beneficiaries hospitalized with PUD
(14). To describe patterns of care, claims data from Medi-
care beneficiaries with a principal diagnosis of PUD hospi-
talized from 1992–1997 were analyzed, as were clinical data
abstracted from hospital discharge records. Using these
data, we sought to answer several questions: What were the
overall outcomes as determined from the Medicare inpatient
claims data during the study period? Did endoscopy proce-
dures during hospitalization influence the outcomes? Based
on abstracted clinical data, what factors are associated with
AJG – January, 2000 Outcomes in Medicare Patients With PUD 265

length of stay and readmission of patients hospitalized with
peptic ulcer disease?
MATERIALS AND METHODS
We performed analyses on two data sets, the annual claims
data and the clinical abstracted data. The claims were se-
lected by the principal diagnoses under various Diagnosis
Related Groups (DRGs), as stated in Appendix A. Descrip-
tive statistics obtained from the claims dataset contained all
of the Medicare inpatient claims from 1992–1997 with
principal diagnoses of peptic ulcer. After excluding the
transfer cases, 867,396 records with 27 variables provided
the following information: Medicare number, age, gender,
race, beneficiary state and county codes, admission date,
discharge date, length of stay, reimbursement amount, prin-
cipal diagnosis codes, procedure codes, DRGs, and dis-
charge destination. The general information was first pro-
filed by year, then by grouping related to endoscopy
procedures. The results were also presented by DRG and
principal diagnosis code.
Clinical data were abstracted from 3067 cases with a
principal diagnosis of gastric or duodenal ulceration (see
Appendix B for the diagnosis codes) selected from a na-
tional sample of Medicare discharges and a cluster sample
from 24 hospital providers having the largest number of
peptic ulcer disease cases in each of five states (Alabama,
Florida, Tennessee, Texas, Louisiana). These samples were
combined to increase the power of the study. The clinical
information was abstracted from charts by the Clinical Data
Abstraction Centers (CDACs)1 from hospital discharges
dated October 1, 1995 to March 31, 1996. The clinical data
were matched and merged with the claims data by Medicare
1
Two CDACs, FMAS Corporation in Columbia, Maryland, and DynKePRO in
York, Pennsylvania, are subcontractors to the Health Care Financing Administration
(HCFA) and abstract data from medical records on their behalf.
identification number, admission, and discharge dates to
obtain readmission rates. There were only 2950 matches out
of the 3067 cases because of the slight changes of identifi-
cation number based on entitlement status. The study on the
readmission rates for the same admission diagnosis included
only these 2950 cases. The Acute Physiology and Chronic
Health Evaluation (APACHE) method was used to estimate
patient health status as a risk factor for the length of stay and
the readmission rates (15). These methods were applied to
assign a risk score for each patient based on the 12 physi-
ological measures, Glasgow Coma Score, age, and chronic
health history. The 12 physiological measures were temper-
ature, blood pressure, heart rate, respiratory rate, oxygen-
ation, arterial pH, CO2, sodium, potassium, creatinine, he-
matocrit, and white blood count. The clinical data contained
all the information except the potassium and the chronic
health history. Missing values were assumed as within the
normal range. To avoid “0” as denominators in the risk
adjustment, “1” was added to the calculated APACHE
scores. The higher the APACHE score, the higher the pa-
tient risk. The risk-adjusted length of stay (LOS) was de-
fined as the LOS divided by the APACHE score. The LOS,
age, and the APACHE scores were continuous variables. All
other variables were binary with value “1”⫽“yes” and
“0”⫽“no.”
The association of LOS and screening or treatment for H.
pylori, screening for NSAID use, and performance of en-
doscopy during hospitalization were studied individually by
applying the two-sample Wilcoxon test. The same studies
were performed by replacing LOS with the risk-adjusted
LOS (LOS/APACHE score). The reason for using these two
methods was to assure the results were reasonable. Using
the risk-adjusted LOS (dividing LOS by APACHE score)
was based on the assumption that the APACHE scores have
effects on the LOS. Treating the APACHE score as an
independent covariant in the multiple regression would test

Table 1. Peptic Ulcer Disease Annual Overall Outcomes (Excluding Transferred Cases)
Average No. of No. of
Year of LOS Reimbursement No. of 30-Day 180-Day No. of No. of
Discharge Discharged (Days) ($) Deaths Readmissions Readmissions Procedures EGD
1992 148,895 7.92 5256 7115 3076 6787 132,827 109,167
1.34% 4.8% 2.1% 4.6% 89.2% 73.3%
1993 147,221 7.49 5407 7005 3154 6615 132,156 111,061
1.32% 4.8% 2.1% 4.5% 89.8% 75.4%
1994 156,586 7.36 5538 7165 3388 7093 141,703 120,619
1.36% 4.6% 2.2% 4.5% 90.5% 77.0%
1995 143,917 6.63 5662 6486 3220 6513 130,907 112,695
1.23% 4.5% 2.2% 4.5% 91.0% 78.3%
1996 136,241 6.24 5799 5871 3021 6028 124,736 108,826
1.16% 4.3% 2.2% 4.4% 91.6% 79.9%
1997 134,536 5.97 5909 5770 2823 4883 124,519 109,571
1.10% 4.3% 2.1% 3.6% 92.6% 81.4%
Total 867,396 6.97 5586 39,412 18,682 37,919 786,848 671,939
1.25% 4.5% 2.2% 4.4% 90.7% 77.5%
LOS ⫽ average length of stay; no. of deaths ⫽ number discharged due to death; 30-day readmissions ⫽ number of readmissions within 30 days with PUD diagnoses; no. of
180 readmissions ⫽ number of readmissions within 180 days with PUD diagnoses; no. of proc ⫽ number of cases having any procedures billed as invasive; no. of EGD ⫽ number
of cases having endoscopies.
266 Xu et al. AJG – Vol. 95, No. 1, 2000

Table 2. Effects of Endoscopy: 6-Year Total (1992–1997) The average patient age in each year was 75 yr. The
Average Average average length of stay decreased 1.95 days (7.92 days in
Number LOS Reimburse- Mortality 1992 to 5.97 days in 1997). The average reimbursement
Claims Group Discharged (Days) ment ($) Rate (%) increased gradually, from $5256 in 1992 to $5909 in
Had procedures 786,848 7.22 5841 4.7 1997. The inpatient mortality rates decreased slightly,
Had endoscopies 671,939 6.51 4643 3.1 from 4.8% in 1992 to 4.3% in 1997. The readmission
Had procedure, 114,910 11.39 12,849 13.7 rates remained stable. The 180-day readmission rate in
no endoscopy
1997 was lower than in other years because of the un-
No endoscopy 195,457 8.56 8832 9.5
availability of 1998 data to compute all the readmissions
LOS ⫽ length of stay.
in 1997. The procedure rates increased 3.4% (89.2% in
1992 to 92.6% in 1997). The endoscopy procedure rates in-
the effects of the APACHE scores and other factors simul- creased 8.1% (73.3% in 1992 to 81.4% in 1997) (Table 1).
taneously. The General Linear Model (GLM) was chosen to
model the log-transformed LOS and detect the joint effects Effects of Endoscopy Procedures
of all the risk factors (the four care measures, age, PUD During Hospitalization (1992–1997)
history, bleeding history, gender, APACHE score). The The total number of Medicare peptic ulcer claims (1992–
association of readmissions and the processes of care were 1997) with any procedures was 786,848. The cumulative
studied by using Fisher’s exact test (16). Finally the logistic statistics for the 6 yr revealed the average length of stay was
regression was performed to investigate the readmissions 7.22 days, the average reimbursement $5841, and the mor-
versus the risk factors (H. pylori positive, treated for H. tality rate 4.7%.
pylori, screened for NSAID, endoscopy performed, age, The total number of Medicare peptic ulcer claims (1992–
gender, PUD history, bleeding history, APACHE score) 1997) with endoscopy procedures was 671,939. The cumu-
(17–20) The readmission rates were not adjusted by the lative statistics for the 6 yr revealed the average length of
APACHE scores because Fisher’s exact test was used and stay was 6.518 days, the average reimbursement $4643, and
the readmission variable required a binary variable. the mortality rate 3.1%.
The total number of Medicare peptic ulcer claims (1992–
1997) with procedures but without endoscopies was
RESULTS
114,910. The cumulative statistics for the 6 yr revealed the
Results of Claims Data Analysis average length of stay was 11.39 days, the average reim-
The total number of peptic ulcer discharges for Medicare bursement $12,849, and the mortality rate 13.7%.
beneficiaries was 867,396 from 1992 to 1997 (Table 1). The total number of Medicare peptic ulcer claims (1992–
The number of claims in 1994 was the highest (156, 586), 1997) without endoscopy procedures during hospitalization
about 16% more than in 1997. The peptic ulcer claims was 195,457. The cumulative statistics for the 6 yr revealed
range from 1.1–1.36% of the total Medicare claims dur- the average length of stay was 8.56 days, the average reim-
ing the 6 yr. The total reimbursement for peptic ulcer bursement $8832, and the mortality rate 9.5% (Table 2).
claims in this 6-yr period was estimated at $4.8 billion. Table 3 indicates that the cases with endoscopy proce-

Table 3. Cases With EGDs Vs Cases Without EGDs (Excluding Transferred Cases)
Cases With EGDs Cases Without EGDs
No. of No. of No. of No. of
Average 30-Day 180-Day Average 30-Day 180-Day
No. Reimburse- No. of Re- Re- No. Reimburse- No. of Re- Re-
Year Discharged LOS ment ($) Deaths admissions admissions Discharged LOS ment ($) Deaths admissions admissions
1992 109,167 7.3 4267 3500 2306 5065 39,728 9.64 7973 3615 770 1722
3.2% 2.1% 4.6% 9.1% 1.9% 4.3%
1993 111,061 6.98 4447 3671 2461 5113 36,160 9.04 8355 3334 693 1502
3.3% 2.2% 4.6% 9.2% 1.9% 4.2%
1994 120,619 6.96 4577 3750 2695 5597 35,967 8.72 8761 3415 693 1496
3.1% 2.2% 4.6% 9.5% 1.9% 4.2%
1995 112,695 6.26 4741 3496 2602 5152 31,222 7.97 8989 2990 618 1361
3.1% 2.3% 4.6% 9.6% 2.0% 4.4%
1996 108,826 5.88 4864 3196 2502 4929 27,415 7.7 9513 2675 519 1099
2.9% 2.3% 4.5% 9.8% 1.9% 4.0%
1997 109,571 5.61 4966 3252 2366 4053 24,965 7.57 10,050 2518 457 830
3.0% 2.2% 3.7% 10.1% 1.8% 3.3%
All 671,939 6.51 4643 20,865 14,932 29,909 195,457 8.56 8832 18,547 3,750 8010
3.1% 2.2% 4.5% 9.5% 1.9% 4.1%
For explanation of parameters measured, see Table 1. EGD ⫽ esophagogastroduodenoscopy.
AJG – January, 2000 Outcomes in Medicare Patients With PUD 267

Table 4. LOS in Quality Measure Groups*

No. of Mean Risk-Adjusted
Parameter Class Cases LOS LOS† 1. p Value‡ 2. p Value‡
Screened or treated for H. pylori No 1431 7.34 0.77 0.0001 0.0009
Yes 1636 6.38 0.71
Screened for NSAIDs No 1001 7.42 0.85 0.0001 0.0001
Yes 2066 6.54 0.68
Assessed for PUD-related risk No 813 6.28 0.72 0.0039 0.8736
Yes 2109 6.87 0.73
EGD performed No 444 7.71 0.84 0.0005 0.0011
Yes 2623 6.68 0.72
* 1. Hypothesis H0 ⫽ same location of LOS between class; 2. Hypothesis H0 ⫽ same location of risk-adjusted LOS between class.
† Risk adjusted LOS ⫽ LOS divided by APACHE Score.
‡ The p values were based on the Wilcoxon rank tests.
LOS ⫽ length of stay; NSAIDs ⫽ nonsteroidal antiinflammatory drugs; PUD ⫽ peptic ulcer disease; EGD ⫽ esophagogastroduodenoscopy; APACHE ⫽ Acute Physiology
and Chronic Health Evaluation.

dures consistently had a shorter average length of stay, Outcomes by DRG

lower reimbursement amount, and lower mortality rate over
the 6 yr. These results were from the claims data without any
risk adjustments.
Outcomes by Principal Diagnosis
The distribution of all the annual statistics by principal
diagnosis followed a similar pattern over the 6 yr (1992–
1997). The top 10 most frequently occurring principal di-
agnoses were almost in the same order for each year. The
top principal diagnosis 531.40 (chronic stomach ulcer with
hemorrhage) covered more than 26,000 claims annually,
with the average length of stay (LOS) decreasing from 7.25
in 1992 to 5.33 days in 1997 and the mortality rate ranging
from 3.5% to 2.6% annually. The patients with principal
diagnosis 532.50 (chronic duodenal ulcer with perforation
without mention of obstruction) had the longest average
LOS (⬎14 days), the largest average reimbursement
(⬎$16,000), and the highest mortality rate (⬎20%) each
year.
The following principal diagnoses had high mortality
rates, on average, annually: 532.00 (duodenal ulcer with
hemorrhage; ⬎5.1%), 533.40 (chronic peptic ulcer with
hemorrhage; ⬎5.1%), and 532.50 (chronic duodenal ulcer
with perforation without mention of obstruction; ⬎20%).
The DRG distribution for peptic ulcer claims of the Medi-
care population (1992–1997) followed almost an identical
pattern as principal diagnosis codes. The top eight most
frequently occurring DRGs were exactly in the same order
for each year. The top frequency DRG 174 (GI hemorrhage
with complication/comorbidity) covered ⬎96,000 claims
annually, with the average LOS ranging between 5.06 and
6.75 days and the mortality rates decreasing from 3.6% in
1992 to 2.8% in 1997. DRGs 178 (uncomplicated peptic
ulcer without complication/comorbidity) and 155 (stomach,
esophageal, and duodenal procedures age ⬎17 yr without
complication/comorbidity) occurred in patients who were
on average 3– 4 yr younger than the overall average age of
75 yr. The following DRGs had mortality rates ⬎10%
annually: 483 (tracheostomy except for face, mouth, and
neck diagnosis; 45.7%–55.5%), 148 (major small and large
bowel procedures with complication/comorbidity; 20.8%–
21.9%), 154 (stomach, esophageal, and duodenal proce-
dures age ⬎17 yr with complication/comorbidity;17.8%–
20.3%), and 170 (other digestive system operating room
procedure with complication/comorbidity; 10.8%–14.5%).
DRG 176 (complicated peptic ulcer) had the highest 180-
day readmission rates: 8.7% in 1992, 7.8% in 1993, 7.8% in
1994, 7.5% in 1995, and 6.5% in 1996. (The percentage in

Table 5. The General Linear Model: Log-Transformed LOS Vs Risk Factors

Standard 95% CI 95% CI
Parameter Estimate Error (Lower) (Higher) Significant p Value
Screened or treated for H. pylori ⫺0.0735 0.0212 ⫺0.1149 ⫺0.0320 * 0.0005
Screened for NSAID usage ⫺0.0705 0.0216 ⫺0.1129 ⫺0.0282 * 0.0011
Risk assessment 0.0381 0.0216 ⫺0.0044 0.0805 0.0787
EGD performed ⫺0.0296 0.0304 ⫺0.0892 0.0300 0.3303
Age 0.0022 0.0001 0.0020 0.0024 * 0.0232
Gender 0.0016 0.0201 ⫺0.0377 0.0410 0.9355
PUD history 0.0329 0.0212 ⫺0.0086 0.0745 0.1206
GI bleeding history ⫺0.0524 0.0227 ⫺0.0970 ⫺0.0079 * 0.0211
APACHE score 0.0257 0.0030 0.00199 0.0315 * 0.0001
* The estimate is statistically significant.
CI ⫽ confidence interval; GI ⫽ gastrointestinal; other abbreviations as in Table 4.
268 Xu et al. AJG – Vol. 95, No. 1, 2000

Table 6. Readmission Rates Vs Processes of Care

No. of
Parameter Class Cases % 30 Days % 60 Days % 90 Days % 180 Days % 360 Days
Screened or treated for H. pylori No 1372 2.77 3.43 3.79 4.96 6.49
Yes 1578 1.90 2.53 2.79 4.06 5.39
Screened for NSAID usage No 970 3.30 4.43 4.48 5.98 7.94
Yes 1980 1.82* 2.22* 2.47* 3.74* 4.90*
Assessed for PUD-related risk No 779 2.95 3.34 3.72 5.39 6.55
Yes 2033 2.16 2.95 3.25 4.33 5.90
EGD performed No 427 2.34 3.28 3.98 4.92 6.32
Yes 2523 2.30 2.90 3.13 4.40 5.83
* The readmission rate is statistically significantly different between groups based on Fisher’s exact test.
Abbreviations as in Table 4.

1997 is not complete due to the unavailability of all 1998
data.)
Results of Clinical Data Analysis
Patients screened or treated for H. pylori had about a 1-day
shorter LOS on average.
Patients screened for NSAID usage had about a 0.9-day
shorter average LOS. Patients who had endoscopy (EGD)
performed had about a 1-day shorter LOS on average. The
difference of the LOS in these three processes of care
(screened or treated for H. pylori, screened for NSAID
usage, and EGD performed) was statistically significant
based on all the tests listed in Table 4.
The group of risk factors (H. pylori screening and/or
treatment, NSAID screening, EGD performance, age, gen-
der, history of PUD, history of bleeding, and APACHE
scores) had joint effects on LOS. The General Linear Model
on the Log-Transformed LOS indicated that H. pylori
screening or treatment, NSAID screening, age, history of
bleeding, and APACHE score were independently associ-
ated with LOS, compared with other variables (Table 5).
All the readmission rates were lower for those patients
who had any of the processes of care documented. However,
only the screening for NSAIDs was associated with differ-
ence in readmission rate (Table 6).
The logistic regression suggested that the joint effect of
all the risk factors significantly influenced the 360-day re-
admission rates. In particular, screening for NSAIDs was
associated with a reduced likelihood of readmission, and
patients with history of bleeding were significantly more
likely to have the 360-day readmissions (Table 7).
DISCUSSION
Favorable trends appeared for patients hospitalized with
PUD over the 6-year period from 1992 through 1997. The
percentage of inpatient peptic ulcer claims was stable from
1992–1994, then decreased consistently from 1994 –1997.
Average LOS, blood usage rate, and mortality rate consis-
tently decreased over the 6 yr. One would assume that these
trends were the result of improved physician education,
advanced technology, better beneficiary health awareness,
and the influence of managed care.
Inpatient claims for PUD dropped 14% from 1994 to
1997, also a trend seen for stroke (DRG 014), acute myo-
cardial infarction (DRG 121), and major joint replacement
(DRG 209). In contrast, pneumonia (DRG 089) and con-
gestive heart failure (DRG 127) inpatient claims remained
relatively constant (Table 8). By far most of the inpatient
admissions studied were for gastrointestinal hemorrhage
(DRG 174), and the longest LOS was for chronic duodenal
ulcer with perforation (DRG 532.50). This could have been
predicted because of the usually dramatic presentation and
uncertainty of outcome in patients with gastrointestinal
hemorrhage and the requirement for surgery in the cases

Table 7. The Logistic Regression: Readmission Within 360 Days Vs Risk Factors
Standard 95% CI 95% CI
Parameter Estimate Error (Lower) (Higher) Significance p Value Odds Ratio
H. pylori positive 0.3411 0.2454 ⫺0.1399 0.8221 0.1645 1.407
Screened or treated for H. pylori ⫺0.2733 0.2491 ⫺0.7615 0.2149 0.2727 0.761
Screened for NSAID usage ⫺0.4777 0.165 ⫺0.8011 ⫺0.1543 * 0.0038 0.62
Risk assessment ⫺0.0237 0.1741 ⫺0.3649 0.3175 0.8917 0.977
EGD performed 0.0266 0.2267 ⫺0.4177 0.4709 0.9067 1.027
Age ⫺0.0129 0.00731 ⫺0.0272 0.0014 0.0774 0.987
Gender 0.0371 0.1615 ⫺0.2794 0.3536 0.8181 1.038
PUD history 0.0911 0.1711 ⫺0.2443 0.4265 0.5946 1.095
GI bleeding history 0.6036 0.171 0.2684 0.9388 * 0.0004 1.829
APACHE score 0.0124 0.0226 ⫺0.0319 0.0567 0.5826 1.013
* The estimate is statistically significant.
Abbreviations as in Tables 4 and 5.
AJG – January, 2000 Outcomes in Medicare Patients With PUD 269

Table 8. Comparison of LOS and Number of Discharges for Other DRGs

1993 LOS 1994 LOS 1995 LOS 1996 LOS 1997 LOS
DRG Discharge (Days) Discharge (Days) Discharge (Days) Discharge (Days) Discharge (Days)
014 369,535 9.6 378,077 8.9 391,124 8.2 383,897 7.6 372,489 7.2
089 433,387 7.9 444,550 7.5 451,709 7.0 428,836 6.5 449,718 6.3
121 162,719 8.5 165,791 7.9 166,981 7.3 160,402 6.9 157,779 6.6
127 692,220 7.1 699,171 6.6 706,241 6.1 688,739 5.8 700,800 5.6
209 308,653 8.4 330,602 7.4 346,926 6.4 341,937 5.8 341,395 5.4
DRG definitions: 014 ⫽ specific cerebrovascular disorder except transient ischemic attack; 089 ⫽ simple pneumonia and pleurisy ⬎ age 17 yr, w/CC (complication/comorbid
condition); 121 ⫽ circulatory disorders with acute myocardial infarction and cerebrovascular accident, discharged alive; 127 ⫽ heart failure and shock; 209 ⫽ major joint and
limb reattachment procedures of lower extremity.
LOS ⫽ length of stay; DRG ⫽ diagnosis-related group.

with perforation. Mortality in excess of 10%, as expected,
was associated with complications and other surgical inter-
vention, such as tracheostomy.
Length of stay in the hospital did decline by 1.95 days
over the study period. The performance of endoscopy was
associated with reduction in LOS. The difference was 2 days
less for cases with endoscopy; however, the nonendoscopy
cases had a much higher mortality rate, suggesting that
unstable patients upon presentation may have perforated or
had a contraindication to endoscopy. Another consideration
to explain the shorter LOS would be early endoscopy for
diagnosis and control of bleeding. Clinical data analysis
revealed a shorter LOS to be statistically related to screening
for H. pylori and separately to screening for NSAID use.
Performance of endoscopy and screening for H. pylori may
be linked, as an opportunity for diagnosis was available if
biopsy could be obtained. Clinical specialty of the admitting
physician was not correlated with the performance of the
processes of care studied. Reduction in LOS has been a
general trend observed for multiple other diagnoses during
the study period (Table 8).
This study found readmission rates to be fairly constant
over the period 1992–1997. One would have predicted a
decrease in readmissions along with the fall in admission
rates and reduction in LOS. Does this imply that H. pylori
is not being eradicated, or that patient education regarding
continued NSAID use is not being heeded or performed?
This question and reasons for regional variation in admis-
sion rates could not be answered from the data in this study.
Also, it confirms one of the limitations in the use of claims
information. Performance of the clinical processes of care
was correlated with readmission rates. Screening for NSAID
use did achieve statistical significance when correlated
with readmission rates at 30, 60, 90, 180, and 360 days
(Table 7). No statistically significant change in readmis-
sion rates was observed in relation to H. pylori investi-
gation or treatment.
The results reported from the Medicare inpatient claims
data were strictly descriptive. It was impossible to draw
conclusions without further clinical investigations or patient
risk adjustments. Nevertheless, the data did serve to increase
the awareness that peptic ulcer disease, although declining
as a cause for admission, continued to influence cost of care,
mortality, and frequency of readmission. The results from
the clinical study suggested that adherence to selected ap-
propriate processes of care is associated with a shorter LOS
and lower readmission rates. Further study of the effect and
mechanisms of these processes in risk-stratified patient sub-
groups is needed.
DISCLAIMER
The analysis upon which this publication is based was
performed under Contract Number 50096P605, entitled
“Utilization and Quality Control Peer Review Organization
for the State of Alabama,” sponsored by the Health Care
Financing Administration, Department of Health and Hu-
man Services. The content of this publication does not
necessarily reflect the views or policies of the Department of
Health and Human Services, nor does mention of trade
names, commercial products, or organizations imply en-
dorsement by the U.S. Government. The author assumes full
responsibility for the accuracy and completeness of the
ideas presented. This article is a direct result of the Health
Care Quality Improvement Program initiated by the Health
Care Financing Administration, which has encouraged iden-
tification of quality improvement projects derived from
analysis of patterns of care. Ideas and contributions to the
author concerning experience in engaging with issues pre-
sented are welcome.
Reprint requests and correspondence: Wen Xu, Ph.D., A.S.A.,
Alabama Quality Assurance Foundation, Suite 200 North, One
Perimeter Park South, Birmingham, AL 35243-3254.
Received Jan. 6, 1999; accepted Sep. 22, 1999.
REFERENCES
1. Marshall BJ, Warren JR. Unidentified curved bacilli in the
stomach of patients with gastritis and peptic ulceration. Lancet
1984;16:1311– 4.
2. Marshall BJ, Armstrong JA, McGechie DB, et al. Attempt to
fulfil Koch’s postulates for pylori campylobacter. Med J Aust
1985;142:436 –9.
3. Goodwin CS, Armstrong JA, Chilvers T, et al. Transfer of
Campylobacter pylori, and Campylobacter mustelae to Heli-
cobacter gen. nov. as Helicobacter pylori comb. nov. and
Helicobacter mustelae comb. nov., respectively. Int J Syst
Bacteriol 1989;39:397– 405.
270 Xu et al. AJG – Vol. 95, No. 1, 2000

4. The Report of the International Update Conference on Heli- APPENDIX

cobacter pylori. American Digestive Health Foundation’s Di-
gestive Health Initiative. May 14, 1997; 1–7. Appendix A. List of the Diagnostic-Related Group (DRG) Codes
5. Walsh JH, Peterson WL. The treatment of Helicobacter pylori and Descriptions (Claims Submitted With Peptic Ulcer Principal
infection in the management of peptic ulcer disease. N Engl Diagnoses)
J Med 1995;333:984 –91.
6. NIH Consensus Development Panel on Helicobacter pylori in DRG Code Description
Peptic Ulcer Disease. Helicobacter pylori in peptic ulcer dis- 148 Major small and large bowel procedures with CC
ease. JAMA 1994;272:65–9. 150 Peritoneal adhesiolysis with CC
7. Soll AH, for the Practice Parameters Committee of Amer- 152 Minor small and large bowel procedures with CC
ican College of Gastroenterology. Medical treatment of 154 Stomach, esophageal, and duodenal procedures
peptic ulcer disease practice guidelines. JAMA 1996;275:
age ⬎17 yr with CC
622–9.
155 Stomach, esophageal and duodenal procedures
8. MedPar data (Medicare billing data), 1996.
9. Graham DY. Helicobacter pylori: Its epidemiology and its role age ⬎17 yr w/o CC
in duodenal ulcer disease. J Gastroenterol Hepatol 1991;6: 157 Anal and stomal procedures with CC
105–13. 170 Other digestive system OR procedures with CC
10. Wilcox CM, Shalek KA, Cotsonis G. Striking prevalence of 172 Digestive malignancy with CC
over-the-counter nonsteroidal anti-inflammatory drug use in 174 Hemorrhage with CC
patients with upper gastrointestinal hemorrhage. Arch Intern 175 GI hemorrhage w/o CC
Med 1994;10:42– 6. 176 Complicated peptic ulcer
11. Armstrong CP, Blower AL. Non-steroidal anti-inflammatory 177 Uncomplicated peptic ulcer with CC
drugs and life threatening complications of peptic ulceration. 178 Uncomplicated peptic ulcer w/o CC
Gut 1987;28:527–32. 460 Nonextensive burns w/o OR procedure
12. Griffin MR, Ray WA, Schaffner W. Nonsteroidal anti-inflam- 468 Extensive OR procedure unrelated to
matory drug use and death from peptic ulcer in elderly per- principal diagnosis
sons. Ann Intern Med 1988;109:359 – 63. 476 Prostatic OR procedure unrelated to
13. Isenberg JI, Soll AH. Peptic ulcer: Epidemiology, clinical principal diagnosis
manifestations, and diagnosis. In: Bennett JC, Plum F, eds. 477 Nonextensive OR procedure unrelated to
Cecil Textbook of Medicine. Philadelphia: W.B. Saunders, principal diagnosis
Co.,1996: 6645– 66. 483 Tracheostomy except for face, mouth, and neck diagnoses
14. Jencks SF, Wilensky GR. The Health Care Quality Improve- CC ⫽ complication/comorbidity; OR ⫽ operating room; GI ⫽ gastrointestinal.
ment Initiative: A new approach to quality assurance in Medi-
care. JAMA 1992;268:900 –3.
15. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. Apache
II: A severity of disease classification System. Crit Care Med
1985;13:818 –29.
16. Fisher LD, van Belle G. Biostatistics, a methodology for the
health sciences. New York: John Wiley & Sons, Inc., 1993. Appendix B. List of Peptic Ulcer Diagnosis Codes and
17. Agresti A. An introduction to categorical data analysis. New Descriptions
York: John Wiley & Sons, Inc., 1996.
Gastric ulcers 531.00–531.91
18. Hettmansperger TP. Statistical inference based on ranks.
Duodenal ulcers 532.00–532.91
Malabar, FL: Krieger Publishing Company, 1991.
Peptic ulcers 533.00–533.91
19. SAS/STAT User’s Guide, Vols. 1, 2. Cary, NC: SAS Institute
Gastrojejunal ulcers 534.00–535.11
Inc., 1990.
Miscellaneous 535.01, 535.11, 535.21, 535.41,
20. Stokes ME, Davis CS, Koch GG. Categorical data analysis
using the SAS system. Cary, NC: SAS Institute Inc., 1995. 535.51, 535.61