Gastroenterology
Volume 136, Issue 1 , Pages 108-114, January 2009

Outpatient Management of Nonvariceal Upper Gastrointestinal Hemorrhage: Unexpected Mortality in Medicare Beneficiaries

Presented in part at Digestive Disease Week, San Diego, California, May 21, 2008, and published in abstract form in Gastroenterology 2008;134:A730.

  • Gregory S. Cooper

      Affiliations

    • Corresponding Author InformationAddress requests for reprints to: Gregory S. Cooper, MD, Division of Gastroenterology, University Hospitals Case Medical Center, 11100 Euclid Avenue, Wearn 247, Cleveland, Ohio 44106-5066. Phone: (216) 844-5386; fax: (216) 983-0347
    • ,
  • Tzyung Doug Kou
    • ,
  • Richard C.K. Wong

Division of Gastroenterology, University Hospitals Case Medical Center, Cleveland, Ohio

Received 19 June 2008; accepted 18 September 2008. published online 26 September 2008.

Article Outline

Background & Aims

Outpatient management of selected patients with nonvariceal upper gastrointestinal hemorrhage (UGIH) has been proposed as a mechanism to decrease resource utilization and expenditures. However, the true prevalence and outcomes of this practice have not been well evaluated in population-based studies.

Methods

We identified a cohort of 9123 episodes of UGIH in 2004 Medicare claims data, including 3506 (38.4%) managed as outpatients. Clinical characteristics, treatment, and outcomes were compared between inpatient and outpatient groups. In order to adjust for potential selection bias in outpatient treatment, propensity score analysis was used to divide patients into quartiles of likelihood for inpatient treatment.

Results

Inpatients tended to be older, with higher comorbidity scores, and were more likely to have a bleeding ulcer or tear. Inpatients were also more likely to undergo endoscopy, including early endoscopy and therapeutics, and require surgery. The overall 30-day mortality rate was 8.0% in the inpatient group and 6.3% in the outpatient group (P < .001), and in the quartile of patients most likely to be managed as inpatients, the 30-day mortality rate was higher in outpatients than in inpatients.

Conclusions

The prevalence of outpatient management of UGIH in the Medicare population was almost 40%, and although patients were likely selected for outpatient management based on clinical criteria, the overall mortality rate in outpatients was considerable. Any potential financial benefit should be balanced against significant mortality rates, at least some of which could possibly be avoided with hospitalization. More optimal selection of candidates for outpatient therapy is likely needed.

Abbreviations used in this paper: CPT-4, Current Procedural Terminology, 4th Edition, EGD, esophagogastroduodenoscopy, ICD-9-CM, International Classification of Diseases, 9th Revision, Clinical Modification, UGIH, upper gastrointestinal hemorrhage

 

See Laine L et al on page 33 in CGH.

Nonvariceal upper gastrointestinal hemorrhage (UGIH) represents a major cause of morbidity, mortality, and health care expenditures.1, 2, 3 In addition to efforts to optimize the timing and type of endoscopic therapeutics in patients with active bleeding and/or endoscopic stigmata of recent hemorrhage, there is an interest in optimizing care for the even larger subset of patients at low risk for adverse outcomes. Two strategies that have been evaluated in the low-risk population include the use of early endoscopy to triage patients4, 5 and the use of outpatient management in selected patients.6

Originally studied by investigators from Kaiser Permanente,6 outpatient management of selected patients with UGIH has been proposed as a more cost-effective strategy for patients in whom endoscopic and clinical findings place them in a low-risk group. Although several studies have compared outcomes of patients treated with inpatient versus outpatient management, all have been limited to either a single hospital or health care network.6, 7, 8, 9, 10, 11, 12 The prevalence and outcomes of outpatient management have not been evaluated in a US population-based sample. We therefore conducted the present study in a cohort of Medicare beneficiaries to obtain national-level data about this practice.

Back to Article Outline

Methods 

Data Sources 

Sample 

The study cohort was obtained from 2004 Medicare claims data. The MEDPAR files contain data from claims for services provided to hospitalized beneficiaries, with each record containing demographic data and up to 10 diagnoses and 10 procedures coded by the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM). The Physician Supplier files include claims from physicians, as well as from other clinicians, and claims from free-standing ambulatory surgery centers. These files contain demographics, 1 primary and up to 3 secondary diagnosis codes (ICD-9-CM), and up to 13 procedures coded according to Current Procedural Terminology, 4th Edition (CPT-4). The Outpatient files include claims from institutional outpatient providers, including hospital outpatient departments, rural health clinics, and renal dialysis facilities. Data elements include demographic characteristics, up to 10 diagnosis codes according to ICD-9-CM, and 1 procedure code according to CPT-4. The Medicare Denominator file contains enrollment data on all Medicare beneficiaries, and was used to track mortality as well as to exclude patients who were not enrolled in Medicare Part B or fee-for-service plans. For this study, given the vast size of these files and the high incidence of UGIH, a 5% random sample of Medicare claims was used. Finally, because individual-level socioeconomic data are not contained in Medicare claims, the 2004 Area Resource File was used. The Area Resource File includes county-level measures of income, educational level, and urban/rural status, as well as county-level health care resources, including density of hospital beds, primary care physicians, and gastroenterologists.

The sample was obtained from all beneficiaries aged 66 years and older in the MEDPAR, Outpatient, or Physician Supplier files with a diagnosis of UGIH as defined by one of the following ICD-9-CM codes: Mallory–Weiss tear (530.7), esophageal hemorrhage (530.82), gastric ulcer with hemorrhage (531.00, 531.01, 531.20, 531.21, 531.40, 531.41, 531.60, 531.61), duodenal ulcer with hemorrhage (532.00, 532.01, 532.20, 532.21, 532.40, 532.41, 532.60, 532.61), peptic ulcer disease, unspecified with hemorrhage (533.00, 533.01, 533.20, 533.21, 533.40, 533.41, 533.60, 533.61), gastrojejunal ulcer with hemorrhage (534.00, 534.01, 534.20, 534.21, 534.40, 534.41), gastritis with hemorrhage (535.01, 535.11, 535.21, 535.31, 535.41, 535.51), duodenitis with hemorrhage (535.61), angiodysplasia of the stomach or duodenum with hemorrhage (537.83), and hematemesis (578.0). Because of the high likelihood of incomplete claims data, patients who were enrolled in Medicare managed care plans and patients who were not enrolled in Medicare Part B were excluded. Using the above sample, we identified discrete episodes of care by using the date of the first encounter within a 30-day period. In order to exclude prevalent or false-positive UGIH diagnoses, if a diagnosis appeared only once in the Outpatient or Physician Supplier files and not in MEDPAR, that episode was excluded unless it was associated with an esophagogastroduodenoscopy (EGD).13

Measures 

The presence and severity of comorbid illnesses were measured using a previously validated claims data–based algorithm.14, 15 The relevant MEDPAR, Outpatient, and Physician Supplier files were searched for a diagnosis code of interest, according to ICD-9-CM during the 365-day to 30-day interval (up to 11 months) before the diagnosis date. In order to maximize the true-positive rate of a listed diagnosis, as previously suggested,16 we only included diagnoses that were either present in MEDPAR or appeared more than once in the Outpatient or Physician Supplier files. A weighted index that combined inpatient and outpatient diagnoses was then used to assign a comorbidity score to each patient.14

The episodes of UGIH identified in the database were divided into inpatient versus outpatient management. Patients who appeared in the MEDPAR file before any outpatient encounters for UGIH or had an initial outpatient encounter 1 day or less before hospitalization were classified as inpatient management. Patients with UGIH not present in the MEDPAR file within 30 days of the index date were classified as outpatient management alone. For the inpatient population, the index date represented the hospital admission date. For outpatients or outpatients who were admitted, the index date was the first occurrence of any medical encounter with the appropriate diagnosis codes. Subsequent hospitalizations within 30 days of discharge were considered to be readmissions.

The use of EGD was identified through the use of ICD-9-CM procedure codes and CPT-4 codes. In addition, procedures were divided into diagnostic examinations (including biopsy) (ICD-9-CM: 45.13, 45.16; CPT-4: 43234, 43235, 43239) and EGD with control of hemorrhage (therapeutic EGD) (ICD-9-CM: 44.43; CPT-4: 43255). The interval between the initial date in the episode of care (as defined above) and the date of the first EGD was calculated. Based on previous work,5 differences of 1 day or less were categorized as “early endoscopy.”

Outcomes 

For each episode of care, 30-day mortality rates were calculated using the Medicare Denominator file and were based on the initial encounter date or date of hospitalization, whichever was earlier. For each episode of care that included hospitalization, the length of stay was calculated based on data from the MEDPAR files. The use of surgery in the treatment of UGIH or its sources was included as an outcome of interest during the episode of care. Performance of surgery was measured through the MEDPAR, Outpatient, and Physician Supplier files for the presence of one or more of the following codes during the episode of care: suture of bleeding lesion (ICD-9-CM: 43.5, 436.6, 43.7; CPT-4: 43501, 43502), partial gastrectomy (ICD-9-CM: 43.5, 43.6, 43.7; CPT-4: 43631, 43632, 43633, 43634, 43635, 43638, 43639, 43810, 43820, 43825), and/or vagotomy (ICD-9-CM: 44.00, 44.01, 44.02, 44.03; CPT-4: 43640, 43641, 43651, 43652).

Analysis 

The analysis was conducted at the level of an episode of care. We compared the characteristics of patients treated with inpatient versus outpatient management, including demographics, comorbidity scores, and census tract measures of socioeconomic status and physician supply. In addition, outcomes including source of hemorrhage, surgery, length of stay, and 30-day mortality were compared between inpatient and outpatient management.

Because of the high likelihood of selection bias in the decision to hospitalize patients with UGIH, we attempted to statistically adjust for this bias using a technique known as propensity scores.17 A propensity score estimates the probability of assignment to a specific treatment given a set of variables, and patients are typically placed into strata based on this probability, independent of outcome such as mortality. For example, predictors of inpatient management could include older age, higher comorbidity scores, and less access to specialists. Moreover, since measured and unmeasured variables are thought to potentially track together, omission of a variable that could not be measured in these data sets would be expected to be associated with less bias with this technique compared with traditional regression models. A stratum with the highest proportion of inpatient management would therefore have a different composition of measured and unmeasured factors than one with the lowest frequency of inpatients. If one evaluated outcomes in individual strata, it would allow comparison in more evenly matched groups of inpatients and outpatients. To create a propensity score, potential predictor variables regardless of statistical significance were included in a multivariate logistic regression model to calculate a probability of inpatient management. These variables were based on previous work6, 7, 8, 9, 10, 11, 12, 18, 19, 20, 21 and included age, gender, race, comorbidity score, state of residence, and county-level measures of socioeconomic status and physician supply. Because we were specifically interested in data available at the time of presentation, the source of hemorrhage was not included. Episodes were then divided into quartiles of likelihood of inpatient management. The associations between inpatient management and 30-day mortality and surgical intervention were evaluated in each propensity quartile.

To test the robustness of the findings, 2 additional models were included. First, a propensity score that included source of hemorrhage as a predictor variable was constructed, and outcomes were measured in each quartile. The rationale for this secondary analysis is that although clinicians would not know the bleeding site at the time of initial assessment, different sources are associated with differences in the clinical magnitude of hemorrhage. The second model was restricted to patients with bleeding peptic ulcer, a cohort in whom the diagnostic and management criteria are more standardized.

Back to Article Outline

Results 

We identified a total of 9123 episodes of care in 2004, including 5617 (61.6%) managed as inpatients and 3506 (38.4%) treated with outpatient management. The latter group included 135 patients who were subsequently admitted to the hospital. The mean age was 78.2 ± 0.8 years, and most patients were female and white and had comorbidity scores of 0 (Table 1). The prevalence of outpatient management varied among the states, from 18.7% to 45.3%.

Table 1. Characteristics of Patients With Outpatient (n = 3506) vs Inpatient (n = 5617) Management
Characteristic% Outpatient% InpatientP value
Age (y) .001
65–6920.515.4
70–7420.917.5
75–7922.122.2
80–8419.621.8
≥8516.923.1
Male42.043.8.09
Race .01
White84.086.3
Black10.59.3
Other5.64.4
Comorbidity index .001
053.554.6
125.012.7
210.811.8
≥310.720.9
Income quartile .56
119.018.3
219.020.0
320.120.9
421.020.3
Missing20.920.5
Education quartile .04
119.317.1
218.720.6
320.620.9
420.520.9
Missing20.920.5
PCP quartile .69
121.021.4
220.720.9
321.021.3
420.721.0
Missing16.615.4
Gastroenterology quartile
120.421.6.08
219.521.3
321.921.0
421.620.7
Missing16.615.4
Hospital bed quartile .59
120.321.2
221.221.0
321.121.2
420.821.2
Missing16.615.4
Source .001
Ulcer/tear25.746.8
Gastritis/duodenitis33.324.3
Other40.931.8

PCP, primary care physician.

Overall, one or more endoscopies were performed in 68.4% of patients, including 81.7% of inpatient episodes and 47.0% of outpatient episodes (P < .001) (Table 2). Both early endoscopy and endoscopic therapeutics were also performed more frequently in inpatients. The most common sources of hemorrhage were gastritis (25.5%), gastric ulcer (20.0%), duodenal ulcer (13.0%), and angiodysplasia (5.9%). Compared with patients managed as outpatients, inpatients were older, had higher comorbidity scores, and were more likely to have an ulcer or Mallory–Weiss tear as a source of hemorrhage (Table 1). In addition, with the exception of minor differences in educational level, county-level measures of socioeconomic status and physician supply were comparable between inpatients and outpatients.

Table 2. Treatment Characteristics and Outcomes of Outpatient and Inpatient Management
Characteristic% Outpatient% InpatientP value
EGD47.081.7.001
Early EGD39.756.4.001
Therapeutic EGD6.819.3.001
Early therapeutic EGD5.413.4.001
LOS (days, Q1–Q3)4 (2,7)4 (3,7).32
Surgery0.21.3.001
30-day mortality6.38.0.001

EGD, esophagogastroduodenoscopy; LOS, length of stay.

The length of hospital stay was similar among both inpatients and outpatients who were admitted, and surgery was performed in 1.3% of inpatients and 0.2% of outpatients (P < .001). Readmission occurred within 30 days in 12.6% of inpatients, and 0.3% of outpatients were later admitted to the hospital. The 30-day mortality rate was 8.0% in the inpatient group and 6.3% in the outpatient group (P < .001) (Table 2).

All demographic, clinical, and small area measures, as well as state of residence, were then entered into a multivariable logistic model to calculate the propensity score or probability of inpatient management. Patients were divided into quartiles of propensity for inpatient management, and the proportion of patients with hospitalization in each quartile was 45.9%, 59.5%, 66.8%, and 74.1%. Outcomes in each of the quartiles are shown in Table 3.

Table 3. Outcomes of Patients in Each Quartile of Propensity for Inpatient Management
% Outpatient% InpatientP value
30-day mortality
Quartile 1 (most inpatient)12.18.4.01
Quartile 26.29.9.003
Quartile 34.06.6.008
Quartile 4 (most outpatient)6.06.4.68
Surgery
Quartile 1 (most inpatient)0.41.1.19
Quartile 20.11.5.002
Quartile 30.21.1.012
Quartile 4 (most outpatient)0.21.4.001

In the quartile of patients most likely to be managed as inpatients, the 30-day mortality rate was higher in the outpatient than in the inpatient group (12.1% vs. 8.4%, P = .01) (Table 3). In the other 3 quartiles, mortality was higher in the inpatient group, although in the quartile with the lowest propensity for inpatient therapy, the difference did not reach statistical significance. In each of the 4 propensity quartiles, the frequency of upper gastrointestinal surgery was higher in the inpatient group, although the rates in the quartile with the highest likelihood of inpatient management did not reach statistical significance (Table 3).

In a secondary analysis, we included the source of hemorrhage as a covariate in the propensity model and created quartiles with a likelihood of inpatient management of 40.5%, 53.1%, 67.9%, and 84.8%. With one exception, in each quartile, the 30-day mortality rates for inpatient and outpatient management were statistically similar, and in contrast to the analysis that excluded source, there was no quartile with a quantitatively higher outpatient mortality (Table 4). Also, after inclusion of source of hemorrhage, surgery was only performed in outpatients in the quartile with the highest likelihood of inpatient management, which is consistent with the higher proportion of bleeding ulcers in this stratum (Table 4).

Table 4. Outcomes of Patients in Each Propensity Quartile Including Source of Hemorrhage as a Predictor Variable
% Outpatient% InpatientP value
30-day mortality
Quartile 1 (most inpatient)7.28.3.49
Quartile 26.77.2.67
Quartile 35.28.7.001
Quartile 4 (most outpatient)6.78.0.24
Surgery
Quartile 1 (most inpatient)2.02.2.86
Quartile 20.00.5.05
Quartile 30.01.2.001
Quartile 4 (most outpatient)0.00.7.003

Finally, we repeated the analysis for only patients with peptic ulcer hemorrhage, including 2462 treated as inpatients and 902 as outpatients. In this cohort, the mortality rates in the inpatient and outpatient groups were 7.6% and 4.6%, respectively (P = .002). In 3 of the 4 propensity quartiles, there was no difference in mortality between inpatients and outpatients, and there was no quartile with higher outpatient mortality rates (Table 5). Rates of surgery also tended to be higher in the inpatients in each quartile.

Table 5. Outcomes of Patients With Peptic Ulcer Hemorrhage in Each Quartile of Propensity for Inpatient Management
% Outpatient% InpatientP value
30-day mortality
Quartile 1 (most inpatient)7.98.7.81
Quartile 26.79.0.33
Quartile 32.86.3.03
Quartile 4 (most outpatient)3.95.2.37
Surgery
Quartile 1 (most inpatient)2.02.2.91
Quartile 20.62.7.11
Quartile 30.82.6.09
Quartile 4 (most outpatient)0.52.0.07

Back to Article Outline

Discussion 

Despite advances in the medical management of peptic ulcer disease, recognition of risk factors for nonsteroidal anti-inflammatory drug toxicity, and eradication of Helicobacter pylori, nonvariceal UGIH remains a prevalent condition. It was estimated that in 2006, there were more than 220,000 hospitalizations, with average hospital charges of approximately $25,000.1 Almost 60% of these admissions were in Medicare beneficiaries.1 Because of comorbid conditions, endoscopic findings, and/or the need for transfusion, many patients require inpatient management; however, a subset of patients who are deemed low risk for adverse outcomes could potentially be managed without hospitalization. Using data from a large number of Medicare enrollees, we have demonstrated that the prevalence of outpatient management is almost 40%, and as high as 45% in one state. Although it appeared that patients were selected for outpatient management based on clinical criteria, the overall mortality rate in the outpatient group was appreciable and surprisingly high. In order to adjust for potential selection bias in treatment allocation, using propensity score techniques, we also identified a subgroup in which mortality rates with outpatient management were even higher than with hospitalization. This subgroup consisted of older patients with a higher comorbidity burden and sources of hemorrhage that were more likely associated with persistent or recurrent bleeding, or both. This subset in particular would have likely benefited from closer observation and intervention associated with hospitalization. However, we were also unable to identify a subgroup in which mortality was sufficiently low (ie, <1%) to safely warrant outpatient therapy.22 Taken together, the findings suggest that in routine practice, more stringent clinical criteria may be needed to appropriately triage patients for optimal management. These would include the routine use of EGD to risk-stratify patients according to endoscopic predictors of continued hemorrhage, and the use of validated clinical prognostic models to triage care.23, 24

Several cohort studies from single hospitals or health plans have evaluated the outcomes of patients with UGIH assigned to outpatient therapy. In a study from a managed care plan, Longstreth and Feitelberg6 evaluated the outcome of 176 patients assigned to outpatient management and reported rebleeding and hospitalization each in only 1% of cases. Although this study used clinical parameters and not endoscopic findings as a triage tool, all patients did undergo endoscopy at some point in the episode. Outpatient management with criteria based on hemodynamics, clinical findings, laboratory values, and findings on urgent endoscopy has been reported in studies from the Unites States,9 Spain,10 and Italy.11 In addition, these criteria have been used to target patients for same-day hospital discharge.8, 25 Using clinical guidelines, all have reported very low rates of rebleeding, surgery, subsequent hospitalization, and mortality. The findings from our study would suggest that if not done routinely, more judicious criteria to predict adverse outcomes should be incorporated. These would include criteria used in the aforementioned clinical studies, as well as the validated Rockall index23, 24 or variants22 including endoscopic triage. Of note is that in our study, only about 40% of patients in the outpatient group underwent endoscopy within 1 day of presentation, and the finding of more equivalent mortality rates in quartiles after source of hemorrhage was included further suggests the benefit of endoscopy as a prognostic marker. In addition, we have recently demonstrated the use of artificial neural networks as a triage tool to predict stigmata of recent hemorrhage and the need for endoscopic therapy,26 with a negative predictive value of at least 92%. This tool could also potentially be used for risk stratification.

Previous studies have also used Medicare data to define outcomes of patients with UGIH. In a study that used inpatient claims from 1991 with a diagnosis of UGIH caused by peptic ulcer disease, we reported a 30-day mortality rate of 8.5% and upper gastrointestinal surgery in 6.9%.27 Xu and colleagues28 evaluated a 6- year period and found an inpatient mortality rate for diagnoses of bleeding peptic ulcer of 4.5%. However, neither study evaluated the important subgroup of Medicare beneficiaries treated as outpatients.

There are several major advantages of using Medicare claims data for studies of UGIH. First, since they encompass nearly all Americans aged 65 and older, the data are nationally representative and include patients treated in community practice as well as tertiary centers. The ability to link individual beneficiaries over time provides data on readmission to the same hospital as well as to other hospitals, and mortality after hospital discharge. Second, by including the Physician Supplier and Outpatient files, the high prevalence of management of UGIH without hospitalization is captured. Third, even with a 5% sample, the large sample sizes provide sufficient statistical power to detect differences not observed in smaller groups of patients.

Despite the advantages of having a large, national population-based sample, we recognize some study limitations. Because the Medicare data are collected for billing purposes and not research, there are potential concerns about both random and systematic bias in diagnosis and procedure coding accuracy.29 In an analysis of 882 discharges from a regional hospital consortium, compared with the medical record, the sensitivity of individual diagnosis codes for nonvariceal UGIH was evaluated.30 Sensitivities of at least 84% were found for gastric ulcer (87.9%), duodenal ulcer (94.3%), gastritis (84.8%), Mallory–Weiss tear (94.2%), and angiodysplasia (95.2%), with only duodenitis having a lower sensitivity (41.2%). In addition, the sensitivity of coding for upper endoscopy was more than 97%. Raiford et al31 used the Saskatchewan Hospital Services Plan database to examine the positive predictive value for ICD-9 coding in 1762 hospitalized patients with UGIH. Using the hospital record as the “gold standard,” bleeding was confirmed in 88%. The positive predictive value of site-specific codes of gastric ulcer, duodenal ulcer, and peptic ulcer was 92%, 93%, and 83%, respectively. The accuracy of codes for UGIH has not been specifically evaluated in either inpatient or outpatient Medicare data, but in a study of inpatient claims, the sensitivity and positive predictive value for a principal diagnosis of peptic ulcer disease were 92% and 89%, respectively.32

The study entry criteria associated with outpatient management were strict, but were designed to exclude false-positive diagnoses. It is not known what effect a less stringent definition would have had on the observed findings. Conceivably, if stable outpatients with trivial bleeds who did not undergo endoscopy had not been excluded from this cohort, the overall outpatient mortality would have been lower. Alternatively, including the subset of outpatients with comorbid conditions that precluded endoscopy and/or the subset of outpatients with minor bleeding in the setting of other life-threatening illnesses would have increased the overall outpatient mortality. We were also unable to measure the cause of death, which may have been due to the UGIH itself, an exacerbation or development of a new comorbid illness because of the UGIH, or unrelated causes. Because the study was limited to patients aged 65 and older who were enrolled in fee-for-service Medicare plans, the generalizability of the findings to other populations could not be measured. Finally, because claims data lack clinical detail, variables such as vital signs, laboratory values, and endoscopic findings could not be measured. These variables, which therefore could not be included in the propensity score construction, were likely used by clinicians to aid in decision making between inpatient and outpatient therapy. However, we did assess other measurable factors that have been shown to have prognostic value, including age and comorbid illnesses.18, 19, 20, 21, 23, 24

In summary, we found that outpatient management of UGIH was a common practice in Medicare beneficiaries, which could potentially have significant aggregate cost savings. However, this potential financial benefit should be balanced against appreciable mortality rates, at least some of which could possibly be avoided with hospitalization. We would suggest the increased use of triage guidelines that incorporate clinical and endoscopic findings.

Back to Article Outline

References 

  1. Agency for Healthcare Research and Quality. HCUPnet: healthcare cost and utilization project. http://hcupnet.ahrq.gov/Accessed July 18, 2008
  2. Bjorkman DJ, Zaman A, Fennerty MB, et al. Urgent vs. elective endoscopy for acute non-variceal upper-GI bleeding: an effectiveness study. Gastrointest Endosc. 2004;60:1–8
  3. Sandler RS, Everhart JE, Donowitz M, et al. The burden of selected digestive diseases in the United States. Gastroenterology. 2002;122:1500–1511
  4. Barkun A, Bardou M, Marshall JK. Consensus recommendations for managing patients with nonvariceal upper gastrointestinal bleeding. Ann Intern Med. 2003;139:843–857
  5. Spiegel BMR, Vakil NB, Ofman JJ. Endoscopy for acute nonvariceal upper gastrointestinal hemorrhage: is sooner better?. Arch Intern Med. 2001;161:1393–1404
  6. Longstreth GF, Feitelberg SP. Outpatient care of selected patients with acute non-variceal upper gastrointestinal haemorrhage. Lancet. 1995;345:108–111
  7. Lee JG, Turnipseed S, Romano PS, et al. Endoscopy-based triage significantly reduces hospitalization rates and costs of treating upper GI bleeding: a randomized controlled trial. Gastrointest Endosc. 1999;50:755–761
  8. Lai KC, Hui WM, Wong BCY, et al. A retrospective and prospective study on the safety of discharging selected patients with duodenal ulcer bleeding on the same day as endoscopy. Gastrointest Endosc. 1997;45:26–30
  9. Cebollero-Santamaria F, Smith J, Gioe S, et al. Selective outpatient management of upper gastrointestinal bleeding in the elderly. Am J Gastroenterol. 1999;94:1242–1247
  10. Almela P, Benages A, Peiro S, et al. Outpatient management of upper digestive hemorrhage not associated with portal hypertension: a large prospective cohort. Am J Gastroenterol. 2001;96:2341–2348
  11. Cipolletta L, Bianco MA, Rotondano G, et al. Outpatient management for low-risk nonvariceal upper GI bleeding: a randomized controlled trial. Gastrointest Endosc. 2002;55:1–5
  12. Klebl FH, Bregenzer N, Schofer L, et al. Comparison of inpatient and outpatient upper gastrointestinal haemorrhage. Int J Colorectal Dis. 2005;20:368–375
  13. Abraham NS, Cohen DC, Rivers B, et al. Validation of administrative data used for the diagnosis of upper gastrointestinal events following nonsteroidal anti-inflammatory drug prescription. Aliment Pharmacol Ther. 2006;24:299–306
  14. Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chron Dis. 1987;40:373–383
  15. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45:613–619
  16. Klabunde CN, Potosky AL, Legler JM, et al. Development of a comorbidity index using physician claims data. J Clin Epidemiol. 2000;53:1258–1267
  17. Rosenbaum PR, Rubin DB. Reducing bias in observational studies using subclassification on the propensity score. J Am Stat Assoc. 1984;79:516–524
  18. Abraham NS, Hartman C, Castillo D, et al. Effectiveness of national provider prescription of PPI gastroprotection among elderly NSAID users. Am J Gastroenterol. 2008;103:323–332
  19. Lanas A, Perez-Aisa MA, Feu F, et al. A nationwide study of mortality associated with hospital admission due to severe gastrointestinal events and those associated with nonsteroidal anti-inflammatory drug use. Am J Gastroenterol. 2005;100:1685–1693
  20. Thomsen RW, Riis A, Christensen S, et al. Outcome of peptic ulcer bleeding among users of traditional non-steroidal anti-inflammatory drugs and selective cyclo-oxygenase-2 inhibitors. Aliment Pharmacol Ther. 2006;24:1431–1438
  21. Marmo R, Koch M, Cipolleta L, et al. Predictive factors of mortality from nonvariceal upper gastrointestinal hemorrhage: a multicenter study. Am J Gastroenterol. 2008;103:1639–1647
  22. Gralnek IM, Dulai GS. Incremental value of upper endoscopy for triage of patients with acute non-variceal upper GI hemorrhage. Gastrointest Endosc. 2004;60:9–14
  23. Rockall TA, Logan RFA, Devlin HB, et al. Incidence of and mortality from acute upper gastrointestinal haemorrhage in the United Kingdom. BMJ. 1995;311:222–226
  24. Rockall TA, Logan RFA, Devlin HB, et al. Variation in outcome after acute upper gastrointestinal haemorrhage. Lancet. 1995;346:346–350
  25. Moreno P, Jaurrieta E, Aranda H, et al. Efficacy and safety of an early discharge protocol in low-risk patients with upper gastrointestinal bleeding. Am J Med. 1998;105:176–181
  26. Das A, Ben Menachem T, Farooq FT, et al. Artificial neural network as a predictive instrument in patients with acute nonvariceal upper gastrointestinal hemorrhage. Gastroenterology. 2008;134:65–74
  27. Cooper GS, Yuan Z, Rosenthal GE, et al. Lack of gender and racial differences in surgery and mortality in hospitalized Medicare beneficiaries with bleeding peptic ulcer. J Gen Intern Med. 1997;12:485–490
  28. Xu W, Hood HM, Burgess PA. The description of outcomes in Medicare patients hospitalized with peptic ulcer disease. Am J Gastroenterol. 2000;95:264–270
  29. In:  Iezzoni LI editors. Risk adjustment for measuring health care outcomes. Ann Arbor, MI: Health Administration Press; 1994;p. 119–175
  30. Cooper GS, Chak A, Lloyd LE, et al. The accuracy of diagnosis and procedural codes for patients with upper gastrointestinal hemorrhage. Gastrointest Endosc. 2000;51:423–426
  31. Raiford DS, Gutthann SP, 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
  32. Fisher ES, Whaley FS, Krushat WM, et al. The accuracy of Medicare's hospital claims data: progress has been made, but problems remain. Am J Public Health. 1992;82:243–248

 The authors disclose the following: Supported in part by a Research and Outcomes and Effectiveness Award from the American Society for Gastrointestinal Endoscopy. The sponsor had no role in the design, conduct, or reporting of the results.

PII: S0016-5085(08)01699-5

doi:10.1053/j.gastro.2008.09.030

Gastroenterology
Volume 136, Issue 1 , Pages 108-114, January 2009

Article Tools