The economics of upper gastrointestinal bleeding in a US managed-care setting: a retrospective, claims-based analysis

Abstract

Objective: To assess 12-month healthcare resource utilization and costs associated with upper gastrointestinal (UGI) bleeding events.

Methods: Patients hospitalized with a UGI bleeding event were identified in US national health-plan claims data (1999–2003) and propensity matched to control patients without UGI bleeding in the same health plan. Matching criteria included age, gender, index date, Charlson Comorbidity Index score, geographic region, and prior medical utilization.

Results: A total of 9,033 UGI-bleed patients and 579,018 control patients met the inclusion criteria, yielding 4,651 matched pairs. After matching, differences between the UGI bleed and general population cohorts remained for office visits, ER visits, and ER costs during the 6-month baseline period prior to the index date. During the 12 months following the index date, both UGI-related healthcare utilization and total healthcare, medical, and pharmacy costs incurred by the UGI-bleed cohort were significantly greater (p< 0.0001) than those incurred by the general population cohort (mean of $20,405 vs. 3,652), even after excluding the initial hospitalization costs (mean of $11,228 vs. 3,652). Costs were primarily due to inpatient hospitalizations (mean of $13,059 for the UGI-bleed cohort vs. $729 for the general population cohort) and ambulatory services (mean of $4,037 for the UGI-bleed cohort vs. $1,537 for the general population cohort). Sixteen percent of the UGI-bleed cohort had a GI-related hospitalization, and about 40% of total costs occurred after the initial hospitalization.

Conclusions: Patients with UGI bleeds experienced significantly higher (p< 0.0001) 12-month health-resource utilization and costs than patients without UGI bleeds. This study provides empirical evidence of the long-term economic burden associated with UGI bleeding. Interpretation of the results should take into account the lack of available information in claims data that could have an effect on study outcomes, such as particular clinical and disease-specific parameters that are not mitigated by propensity score and comorbidity index matching. In addition, this study is limited by the intensive demographic matching that was done between the two cohorts, which may have eliminated the sickest UGI patients and the healthiest general health-plan population patients.

Key words::

• claims analysis
• costs
• economics
• gastrointestinal disease
• hemorrhage

Introduction

Bleeds of the gastrointestinal (GI) tract have serious morbidity and mortality consequences. Upper gastrointestinal (UGI) bleeding refers to any bleeding or hemorrhage in the UGI tract. The most commonly occurring events are gastric ulcers (21%), varices (21%), duodenal ulcers (13%), and mucosal erosive disease (12%)¹. The incidence rate of acute UGI bleeding is estimated to be 40–150 episodes per 100,000 persons annually, and results in 10,000–20,000 deaths a year². The corresponding mortality rate is estimated at 6–10%^3–5, and can rise to as high as 25% when UGI bleeding is recurrent⁴. Old age is associated with UGI bleeding, and the incidence rate increases 200-fold for someone in their 90s compared to someone in their 30s⁶. Over 40% of UGI patients are 60 years of age or older, and a retrospective study found that 73% of deaths due to UGI hemorrhage occurred in patients over 60⁴. Risk of death from UGI bleeding can be compounded at any age by comorbidities, including chronic renal and hepatic dysfunction, malignancies, coronary artery disease, and chronic obstructive pulmonary disease^3–7.

UGI bleeding imposes $2.5 billion in annual costs to the US healthcare system², and results in more than 300,000 hospital admissions every year⁶. The high rate of hospitalizations is partly due to the fact that patients who don't stabilize after acute bleeding may require ICU admission for treatment of volume depletion and shock^7–9. The annual rate of hospital admissions for UGI bleeding is estimated to be from 36 to 102 per 100,000 persons⁴. However, as practice patterns have evolved, length of hospital stay has decreased over the last 10 years, while endoscopy within 24 hours of hospitalization has increased¹⁰.

UGI bleeds require intensive medical interventions that add considerably to healthcare-resource utilization and treatment cost. However, the total cost of treatments and interventions required after an acute event are often detached from the initial event, masking the full impact on health-resource utilization. Direct and indirect costs of UGI bleeding are compounded by the frequency of events, particularly among patients who are vulnerable to severe UGI bleeds. The total cost of a UGI event (including diagnosis and treatment for anemia or other symptoms) is significant¹¹, but the long-term cost of care following hospitalization for a UGI bleed has not been well documented¹². Therefore, the main objective of this study was to assess the 12-month health-resource utilization costs associated with severe UGI bleeding events requiring hospitalization. This was accomplished by examining the cost of healthcare during the 1-year period following hospitalization and comparing it to the 1-year cost of care in the general health-plan population.

Patients and methods

The objective of this study was to provide information on the economic burden of UGI bleeding; in particular, the incremental cost of healthcare, defined as the difference in healthcare costs during the 1-year period following hospitalization compared to the 1-year cost of care in the general health-plan population. This study defined UGI bleeds or hemorrhages based on specified ICD-9 diagnostic codes (see below).

Data source

This study was a retrospective, administrative claims-based analysis of medical and pharmacy claims data and enrollment information from a large managed-healthcare plan in the US, for services or products provided between July 1, 1999 through December 31, 2004. Claims were submitted by physicians, facilities, and pharmacies for payment of services provided to covered health-plan members. The administrative claims database included data for approximately 14 million covered lives with both medical and pharmacy benefits. The health plan comprised discounted fee-for-service independent practice association (IPA) plans spanning the US, with the largest concentration in the southern and midwestern regions. All study data were de-identified and accessed with protocols compliant with the Health Insurance Portability and Accountability Act (HIPAA).

UGI bleeding study subject identification

Patients at least 18 years of age with evidence of a hospitalization for a UGI bleeding event during the period from 01/01/2000 through 12/31/2003 were identified from the claims data and were selected for inclusion if they had not been hospitalized for UGI bleeding or for UGI-related surgery during the 6-month pre-index (baseline) period. The admission date of the first occurring hospitalization for a UGI bleeding event, without ICD-9-CM diagnostic evidence of trauma (see Table 1), was defined as the index date. Patients who were not continuously enrolled with both medical and pharmacy benefits for at least 6 months prior to the index date (‘baseline’ period) and at least 12 months (or until date of death if the patient died within the first 12 months) following the index date (‘follow-up’ period) were excluded from the study sample.

Table 1. Definition of trauma: International Classification of Diseases, Ninth Revision, Clinical Modification.

ICD-9 codes for trauma
805.xx–809.x	929.x	E910.x
835.xx	933.x	E916–E929.x
839.xx	935.x–938	E950.x–E977
860.x–869.x	954.x	E979.x
874.xx–877.1	956.x	E985.x–E999.x
879.x	958.2
901.x–902.xx	958.5
905.1	959.1x
905.5	E800.x–E849.x

Patients were considered to have been hospitalized for a UGI bleeding event if they had at least one medical claim with an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis code indicating UGI bleeding listed in the primary position of an inpatient hospitalization (ICD-9-CM 456.0, 456.20, 530.7, 530.82, 531.0x, 531.2x, 531.4x, 531.6x, 532.0x, 532.2x, 532.4x, 532.6x, 533.0x, 533.2x, 533.4x, 533.6x, 534.0x, 534.2x, 534.4x, 534.6x). The resulting UGI-bleed cohort consisted of 9,033 patients.

General health-plan population subject identification

A second cohort of patients at least 18 years of age with no evidence of hospitalization for a UGI bleeding event or UGI-related surgery during the 6-month pre-index period was identified from 2 million randomly selected patients enrolled from 01/01/2000 through 12/31/2003. Since these patients did not have a natural index date (e.g. diagnosis date), the index date was based on a randomly selected service date. Random selection was based on a uniform distribution (e.g. probability of a specific date selected for a patient with 12 service dates would be 1/12th). The resulting general population cohort consisted of 579,018 patients.

Matching

The patients in the UGI-bleed cohort were matched 1:1 with patients drawn from a 10% general population sample by hard-matching several patient characteristics and propensity scores, creating 4,651 matched pairs. Patients were hard-matched on the following data: year/quarter (e.g., 2000 Q1) of study entry; age ±2 years; gender; Charlson Comorbidity Index (CCI) score¹³, and propensity score + 0.0001.

Propensity score model

The propensity score is the predicted probability of being a member of the UGI-bleed sample based on a set of observed covariates that are entered into a logistic regression model. In a cohort study, matching the propensity scores of subjects in two cohorts serves to balance all of the observed covariates which may be too numerous to independently hard-match. Factors considered for use in the propensity score model included patient demographics (age, gender), medication use, factors associated with mortality (e.g. comorbid conditions), resource utilization and costs, and time of cohort entry (quarter/year).

Study measures

Variables used in matching

Patient demographic variables, age, gender, and geographic location were captured from the enrollment data. Age was calculated during the year of the index date. Filled prescriptions in the pre-index period were examined for medications of interest from a variety of different classes.

Medications of interest were chosen because they reflect drug classes used by ambulatory patients on a chronic basis, and are in common use and/or may influence the patient's likelihood of incurring UGI bleeding. The included drug categories were: angiotensin-converting enzyme inhibitors; angiotensin-receptor blockers; beta-blockers; alpha-blockers; antiarrhythmics; anticoagulants; antiplatelet agents; calcium channel blockers; digoxin; diuretics; hydralazine; nitrates; beta agonists; anticholinergics; other antihypertensives; lipid-lowering agents; antirheumatics; narcotic analgesics; antidiabetic agents; proton pump inhibitors (PPIs); H₂-receptor antagonists (H₂RAs); misoprostol; sucralfate; selective serotonin reuptake inhibitors; bisphosphonates; nonsteroidal anti-inflammatory drugs (NSAIDs), including but not limited to COX-2 inhibitors; corticosteroids; and vitamin K. Comorbidities, an important confounding factor, were measured during the baseline period using the CCI¹³, a clinical index that incorporates 19 categories of comorbidity, which in their claims-adapted version are defined by ICD-9-CM diagnosis codes¹⁴. Each category is assigned a weight of 1 to 6 to indicate relative comorbidity, which is based on the adjusted risk of 1-year mortality. Patients' CCI score is the weighted sum of the conditions. The overall comorbidity score reflects the cumulative increased likelihood of 1-year mortality; the higher the score, the more severe the comorbidity burden.

Outcome variables

The main outcome variable evaluated in this study was the healthcare utilization and cost outcomes of UGI bleeding reported in claims data in the 12-month period following the index date. Healthcare utilization refers to all medication and services covered by the health plan including, but not limited to, office and outpatient visits, hospitalizations, administered and prescription pharmacy medications, procedures, and laboratory services. In addition, pre-index healthcare costs from medical and pharmacy claims for services delivered in the 6-month pre-index period were calculated as a function of health plan- and patient-paid amounts.

Analysis

The study outcomes (utilization of healthcare resources) were compared between the matched cohorts using a chi-squared test. Comparisons of all continuous variables were based on bivariate t-tests, and a chi-squared test was used for dichotomous variables. All analyses were conducted using the SAS (version 9.0) software program (SAS Institute, Inc., Cary, NC, USA). Matching and propensity score models were performed using Stata (version 9.0) statistical software (Stata Corporation, College Station, TX, USA).

Results

Patient characteristics

Table 2 displays the patient characteristics and baseline distributions on the CCI for the two cohorts prior to matching. A greater percentage of individuals in the UGI-bleed cohort experienced comorbidities, showing a significantly higher mean CCI score (0.92 vs. 0.13; p< 0.0001) when compared to the general population cohort. Following the matching procedure, the CCI score and patient demographic characteristics of the 4,651 matched pairs from the UGI bleed and general population cohorts appear very similar (Table 3). Despite a relatively large sample size, formal statistical tests show no evidence of a statistically significant difference between cohorts. After matching, the vast majority of clinical characteristics of the UGI bleed and general population cohorts also appear to be very similar. Only other antihypertensive (p=0.0360) and PPI (p=0.0234) use showed evidence of a statistically significant difference between cohorts at the 0.05 significance level. In fact, 16 of the 32 comparisons had a p-value greater than 0.50. Furthermore, a lower significance level is often recommended when several comparisons are conducted. Consequently, lowering the significance level to 0.01 yielded no significant differences in clinical characteristics after matching.

Table 2. Pre-match: patient characteristics and baseline Charlson Comorbidity Index score.

Variables	UGI bleed n=9,033 n (%)	General population n=579,018 n (%)	p-value
Age
18–34	837 (9)	142,513 (25)	< 0.0001
35–64	6,225 (69)	408,719 (71)	0.0005
65+	1,971 (22)	27,786 (5)	< 0.0001
Gender
Female	3,942 (44)	319,762 (55)	< 0.0001
Male	5,091 (56)	259,256 (45)	< 0.0001
Comorbidities
AIDS	26 (0.29)	211 (0.04)	< 0.0001
Hemiplegia or paraplegia	53 (0.59)	239 (0.04)	< 0.0001
Cerebrovascular disease	500 (5.54)	3,320 (0.57)	< 0.0001
Diabetes with chronic complications	341 (3.78)	1,582 (0.27)	< 0.0001
Mild or moderate diabetes	1,139 (12.61)	20,885 (3.61)	< 0.0001
Moderate or severe liver disease	166 (1.84)	76 (0.01)	< 0.0001
Mild liver disease	163 (1.80)	648 (0.11)	< 0.0001
Metastatic solid tumor	112 (1.24)	279 (0.05)	< 0.0001
Any malignancy	497 (5.50)	8,787 (1.52)	< 0.0001
Congestive heart failure	545 (6.03)	2,199 (0.38)	< 0.0001
Dementia	41 (0.45)	154 (0.03)	< 0.0001
Myocardial infarction	148 (1.64)	1,512 (0.26)	< 0.0001
Peptic ulcer disease	418 (4.63)	1,111 (0.19)	< 0.0001
Chronic pulmonary disease	1,141 (12.63)	15,265 (2.64)	< 0.0001
Peripheral vascular disease	323 (3.58)	1,613 (0.28)	< 0.0001
Renal disease	313 (3.47)	925 (0.16)	< 0.0001
Rheumatologic disease	239 (2.65)	3,154 (0.54)	< 0.0001
Mean Charlson score (SD)	0.92 (1.51)	0.13 (0.46)	< 0.0001

AIDS, acquired immunodeficiency syndrome; SD, standard deviation; UGI, upper gastrointestinal.

Table 3. Post-match: patient demographics and clinical characteristics.

Variables	UGI-bleed n=4,651 n (%)	General population n=4,651 n (%)	p-value
Age
18–34	622 (13.37%)	609 (13.09%)	0.6908
35–64	3,665 (78.80%)	3,695 (79.45%)	0.4441
65+	364 (7.83%)	347 (7.46%)	0.5071
Gender
Female	1,950 (41.93%)	1,950 (41.93%)	1.0000
Male	2,701 (58.07%)	2,701 (58.07%)	1.0000
Mean Charlson score (SD)	0.14 (0.39)	0.14 (0.39)	1.0000

SD, standard deviation; UGI, upper gastrointestinal.

After matching, differences between the UGI-bleed and general population cohorts remained for office visits, ER visits, and ER costs during the 6-month baseline period prior to the index date (Table 4). More specifically, the UGI-bleed cohort experienced a greater number of ER visits (mean of 0.22 vs. 0.14 visit per patient; p< 0.001) and incurred higher ER costs ($99.95 vs. 51.31; p< 0.001) than the general population cohort. Furthermore, the UGI-bleed cohort experienced significantly fewer office visits (mean of 2.65 vs. 2.84 visits per patient; p=0.003) when compared to the general population cohort. Total healthcare ($1,823 vs. 1,711; p=0.23), medical ($1,356 vs. 1,229; p=0.16), and pharmacy costs ($468 vs. 482; p=0.44) were similar between the cohorts during the baseline period (Table 4).

Table 4. Post-match: baseline healthcare utilization and costs.

Variables	UGI bleed mean (SD) n=4,651	General population mean (SD) n=4,651	p-value
Number of visits
Inpatient	0.05 (0.27)	0.05 (0.25)	0.4963
Outpatient	1.20 (2.16)	1.27 (2.20)	0.1382
Office	2.65 (2.89)	2.84 (3.38)	0.0030
ER	0.22 (0.60)	0.14 (0.58)	< 0.0001
Costs
Total cost	1,823 (5,028)	1,711 (3,989)	0.2302
Medical cost	1,356 (4,852)	1,229 (3,789)	0.1600
Outpatient cost	698 (1,580)	730 (1,542)	0.3138
ER cost	100 (380)	51 (277)	< 0.0001
Inpatient cost	457 (4,127)	383 (3,172)	0.3322
Rx cost	468 (917)	482 (826)	0.4417

Mean ER, Inpatient and outpatient costs do not sum to mean medical cost value due to medical service provided in other settings (e.g. nursing and rehabilitation facilities, home, and ambulance).

ER, emergency room; Rx, prescription; SD, standard deviation; UGI, upper gastrointestinal.

Post-index date outcomes

UGI-related healthcare utilization was significantly greater among the UGI-bleed cohort during the 12-month period following the index date (p< 0.0001) (Table 5). By definition, 100% of the UGI-bleed patients had a UGI-related hospitalization. Among UGI-bleed patients, 16% (723 of 4,651 patients) had a UGI-related rehospitalization within the first year, resulting in 79% of inpatient cost attributable to UGI bleeding. It is noteworthy that the proportion of patients with a non-GI-bleed-related hospitalization was similar between cohorts. For all types of ambulatory visits, the proportion of patients experiencing a visit was greater for the UGI-bleed cohort than for the general population cohort, regardless of physician specialty (general practice, gastroenterology, or other). While the proportion of patients with a retail prescription was statistically greater for the UGI bleed than general population cohort, it should be noted that the difference, while statistically significant, is relatively small for non-GI-related medication use (89 vs. 85%; p< 0.001).

Table 5. 12-month outcomes: healthcare utilization.

Variables	UGI-bleed n=4,651	General population n=4,651	p-value
	n (%)	n (%)
Inpatient hospitalizations
Any UGI-related hospitalizations	4,651 (100.0%)	108 (2.32%)	< 0.0001
Any UGI-related rehospitalizations	723 (15.6%)	NA
Any non-GI-related hospitalizations	324 (6.97%)	304 (6.54%)	0.4085
Ambulatory visits
Any GP ambulatory visits	3,928 (84.45%)	2,928 (62.95%)	< 0.0001
Any GP ambulatory visits related to UGI	2,642 (56.80%)	655 (14.08%)	< 0.0001
Any gastroenterologist ambulatory visits	2,150 (46.23%)	359 (7.72%)	< 0.0001
Any gastroenterologist ambulatory visits related to UGI	1,933 (41.56%)	266 (5.72%)	< 0.0001
Any other ambulatory visits	4,211 (90.54%)	3,363 (72.31%)	< 0.0001
Any other ambulatory visits related to UGI	2,866 (61.62%)	664 (14.28%)	< 0.0001
Retail prescription medication use
Any UGI-related filled prescriptions	3,732 (80.24%)	1,231 (26.47%)	< 0.0001
Any non-GI-related filled prescriptions	4,166 (89.57%)	3,990 (85.79%)	< 0.0001
Diagnostic and therapeutic procedures
Any UGI-related procedures	1,826 (39.26%)	412 (8.86%)	< 0.0001
Any non-GI-related procedures	3,899 (83.83%)	2,965 (63.75%)	< 0.0001
Laboratory tests
Any laboratory services related to UGI	2,497 (53.69%)	1,167 (25.09%)	< 0.0001

ER, emergency room; GI, gastrointestinal; GP, general practitioner; Rx, prescription; UGI, upper gastrointestinal.

The combined health plan- and patient-reimbursed amounts for healthcare services were compared between the UGI bleed and general population cohorts for the 12-month period following the index date. Total healthcare costs, medical costs, and inpatient costs are reported as both including and excluding the initial UGI-bleed hospitalization (Table 6). The initial hospitalization represented a large amount of the total cost of care for the UGI-bleed cohort, increasing the average 1-year cost from $11,228 to $20,404. When including the cost of initial hospitalization, the main cost drivers for the UGI-bleed cohort were inpatient hospitalizations (about 64% of total costs), while outpatient visits and prescription pharmacy accounted for 23% of expenditures (Table 6). When excluding the cost of initial hospitalization, the main cost drivers for the UGI-bleed cohort were inpatient hospitalizations (35% of total costs) and outpatient visits (39% of total costs). Regardless of whether the initial hospitalization was included or excluded, total healthcare, medical, and pharmacy costs were significantly greater (p< 0.0001) among the UGI-bleed patients.

Table 6. 12-month outcomes: healthcare costs (in US dollars).

Variables	Including initial UGI hospitalization			Excluding initial UGI hospitalization
	UGI bleed n=4,651	General population n=4,651		UGI bleed n=4,651	General population n=4,651
	Mean (SD)	Mean (SD)	p-value	Mean (SD)	Mean (SD)	p-value
Total costs	20,405 (37,842)	3,652 (6,927)	< 0.0001	11,228 (25,776)	3,652 (6,927)	< 0.0001
Medical costs	18,721 (37,344)	2,514 (6,399)	< 0.0001	9,544 (25,188)	2,514 (6,399)	< 0.0001
Inpatient costs	13,059 (31,594)	729 (4,269)	< 0.0001	3,882 (18,592)	729 (4,269)	< 0.0001
Ambulatory costs	4,037 (8,501)	1,537 (3,787)	< 0.0001	4,037 (8,501)	1,537 (3,787)	< 0.0001
ER costs	368 (1,158)	113 (522)	< 0.0001	368 (1,158)	113 (522)	< 0.0001
Other medical costs	1,257 (8,586)	136 (709)	< 0.0001	1,257 (8,586)	136 (709)	< 0.0001
Rx costs	1,684 (2,439)	1,138 (1,817)	< 0.0001	1,684 (2,439)	1,138 (1,817)	< 0.0001
Total UGI-related costs	14,662 (31,842)	578 (2,630)	< 0.0001	5,488 (18,757)	578 (2,630)	< 0.0001
Medical UGI-related costs	14,104 (31,776)	438 (2,551)	< 0.0001	4,930 (18,660)	438 (2,551)	< 0.0001
Inpatient UGI-related costs	12,259 (30,645)	199 (2,182)	< 0.0001	3,086 (17,439)	199 (2,182)	< 0.0001
Ambulatory UGI-related costs	1,221 (2,704)	208 (1,036)	< 0.0001	1,221 (2,704)	208 (1,036)	< 0.0001
ER UGI-related costs	195 (658)	21 (179)	< 0.0001	195 (658)	21 (179)	< 0.0001
Other UGI-related medical costs	429 (3,277)	9 (90)	< 0.0001	429 (3,277)	9 (90)	< 0.0001
Rx UGI-related costs	558 (747)	141 (405)	< 0.0001	558 (747)	141 (405)	< 0.0001

ER, emergency room; GI, gastrointestinal; Rx, prescription; SD, standard deviation; UGI, upper gastrointestinal.

Discussion

The objective of this study was to assess the 1-year burden of illness and costs among patients hospitalized for UGI bleeding. Accordingly, the authors identified a sample of individuals who experienced a UGI bleeding event (UGI-bleed cohort) and matched them to a sample of individuals who had not experienced a UGI bleeding event (general population cohort), using a number of specific patient characteristics and their propensity scores. Prior to matching, significant differences existed between the UGI-bleed cohort and the general population cohort on the CCI score (p< 0.0001). After matching, 51% of the eligible UGI-bleed patients were included in the final dataset. The matching methods used in the present study provided a good balance for the vast majority of patient and clinical characteristics and pre-index healthcare utilization. However, after matching, the UGI-bleed cohort still experienced significantly more ER visits (p< 0.0001) and fewer outpatient visits (p=0.1382) in the pre-index period, when compared to the general population cohort.

In the pre-index period, total healthcare costs were similar between the UGI-bleed and general population cohorts ($1,823 vs. 1,711; p=0.23). However, during the post-index period (12-month outcomes, Table 6), the UGI-bleed cohort experienced significantly greater total healthcare costs, including the cost of initial hospitalization, when compared to the general population cohort ($20,405 vs. 3,652; p< 0.0001). The majority of costs were due to inpatient hospitalizations and ambulatory services, particularly the initial hospitalization, which accounted for 45% of the total 1-year costs. Of those patients initially hospitalized for UGI bleed, 16% were rehospitalized during the 1-year follow-up period and incurred an additional $3,882 in inpatient costs. In addition to greater inpatient costs, the UGI-bleed cohort had greater ambulatory costs, when compared to the general population cohort. In the 12 months following the initial hospitalization, the UGI-bleed patients experienced a 516% increase in healthcare costs, compared to a 115% increase in costs for the general population cohort. Notably, 92% of the UGI-bleed patients in this study were under 65 years of age, and 13% were under 35 years. While this may be influenced by the nature of the dataset (a commercially insured population), it is surprising to identify a large proportion of UGI-bleed hospitalizations occurring among a younger population.

This long-term perspective sheds light on the overall healthcare burden of patients with UGI bleeding, and is usually not accounted for in cost-effectiveness and other economic evaluations¹⁵.

The reported findings have important implications for the management and treatment of UGI bleeding. Given the high costs that patients in the UGI-bleed cohort incurred due to rehospitalization and ambulatory services, these results highlight the potential economic benefit to be gained from successful long-term management and prevention of UGI bleeding. Drugs that control acid levels, such as PPIs and H₂RAs, have been shown to be effective at preventing recurrent UGI bleeding^15–17, and COX-2 inhibitors such as celecoxib produce lower rates of GI bleeding, when compared against nonselective NSAIDs. In a double-blind study in which patients with duodenal ulcers were followed up to 3 years, the H₂RA ranitidine was shown to be effective at preventing recurrent bleeding¹⁸. Ranitidine has also been shown to be effective at preventing some types of ulcers that occur in rheumatoid arthritis patients taking NSAIDs¹⁹. Additionally, in a prospective cohort study of patients with UGI bleeding who were followed for a mean of 3 years, use of a PPI was associated with a reduced risk of recurrence²⁰. However, a retrospective cohort study showed that use of a PPI in conjunction with celecoxib did not provide any additional benefit in terms of reduced risk of recurrence, when compared to celecoxib alone, for patients aged 64–75 years²¹.

Cost-effectiveness studies are needed that examine the long-term economic benefits obtained from managing and preventing UGI bleeding with medication. In a modeling study where 1000 hypothetical patients with acute UGI bleeding were followed over a period of 60 days, use of a PPI before endoscopic therapy resulted in a total savings of 20,700 Canadian dollars, when compared to endoscopic therapy alone²².

While claims data are valuable for examining healthcare outcomes, treatment patterns, healthcare-resource utilization, and costs, they are collected for the purpose of payment, not research. Therefore, there are certain limitations associated with the use of claims data. It is possible that information is not available in claims data that could have an effect on study outcomes, such as particular clinical and disease-specific parameters that are not mitigated by propensity score and comorbidity index matching. Furthermore, presence of a diagnosis code on a medical claim is not positive presence of disease, as the diagnosis code may be incorrectly coded. Total health-related utilization costs in the 12-month period following the UGI event may be overestimated because a patient's frequent interaction with the healthcare system could result in more opportunities for additional diagnoses and treatments; conversely, total health-related utilization costs could be underestimated because of mortality selection bias. Intensive demographic matching between the disease cohort and the general population in this commercial database may have eliminated the sickest UGI patients and the healthiest general population patients. As a group, individuals who are enrolled in a commercial health insurance plan may be healthier than an uninsured or government-insured population. Prescription drug costs may be higher in the UGI-bleed cohort because a hospitalization event may augment a patient's compliance and persistence with taking medications.

Conclusions

This study provides empirical evidence of the additional long-term economic burden associated with UGI bleeding. The UGI-bleed cohort had significantly greater healthcare-resource utilization and costs compared with the matched general population cohort, even after excluding the costs associated with the initial hospitalization (p< 0.0001). This was true for both UGI-related and non-related events, excluding the number of non-GI-related hospitalizations which did not significantly differ between the cohorts. Many of the additional costs incurred by the UGI-bleed cohort were GI-related, presumably for follow-up care, but they also incurred greater non-GI-related costs.

These study findings shed light on the overall healthcare burden associated with UGI bleeding, and have important implications for its management and treatment. Although intensive demographic matching between the two cohorts may have excluded the sickest UGI-bleed patients and the healthiest general population patients, the study sample appears to be a fair representation of the broader patient population with UGI bleeding, thus allowing for generalizability of the results. Given the high costs that patients in the UGI-bleed cohort incurred due to rehospitalization and ambulatory services, these results highlight the potential economic benefit to be gained from successful long-term management and prevention of this disorder.

Acknowledgements

Declaration of interest: Pfizer, Inc., provided the funding for this study. CW and BC received honoraria from Pfizer for consulting services on the study; VZ and HH have disclosed that they are employed by i3 Innovus, which had a research contract for this study; and LC and GZ have disclosed that they are employed by Pfizer.