Using the Nationwide Inpatient Sample (NIS) (21), we analyzed hospital discharge data during 1998–2005 for the general US population. The NIS is the largest all-payer, inpatient-care database in the United States; it is produced by the Healthcare Cost and Utilization Project (HCUP), sponsored by the Agency for Healthcare Research and Quality in partnership with public and private statewide data organizations (21,22). The NIS is a stratified probability sample of hospitals in participating states designed to approximate a 20% sample of all US community hospitals. Hospitals in the sampling frame include short-term, nonfederal general and specialty hospitals from as many as 37 states (in 2005). We calculated national hospitalization estimates using discharge weights developed by HCUP.

Diagnosis codes from the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM), were used to define PUD in the following terms: peptic ulcer (533), gastric ulcer (531), gastrojejunal ulcer (534), and duodenal ulcer (532) (23). Since the first-listed diagnosis is the condition chiefly responsible for the hospitalization, hospitalizations in which 1 of these codes was listed as the first diagnosis were considered to be more specific than hospitalizations in which 1 of these codes was listed as 1 of as many as 15 diagnoses. We limited the analysis to rates of hospitalization for first-listed PUD diagnoses, unless otherwise stated. In a separate analysis, we selected records with H. pylori infection as 1 of as many as 15 diagnoses (ICD-9-CM code 041.86), without regard to PUD on the record.

The PUD and H. pylori hospitalizations were examined by age group (<20, 20–44, 45–64, and >65 years of age), sex, race/ethnicity (white, black, Hispanic, and Asian or Pacific Islander), geographic region (Northeast, Midwest, South, and West), designated ulcer type (peptic, gastric, gastrojejunal, and duodenal), other diagnoses listed along a PUD diagnosis, and procedures. Race/ethnicity was missing in the record for 26.0% of hospitalizations.

We selected first-listed hospitalizations for gastritis/duodenitis (ICD-9-CM code 535) as a comparison group to ensure that a change in the PUD hospitalization rate was not attributable to changes in diagnoses resulting from the increased specificity associated with endoscopy. Hospitalizations for all diagnoses were also examined as a comparison group to ensure that a change in hospitalizations for PUD was not merely a reflection of a change in the total number of hospitalizations for all diagnoses.

Annual average hospitalization rates were expressed as the number of hospitalizations per 100,000 population. The hospitalization rates were calculated by using the weighted number of hospitalizations and the census population for each year of the study period from HCUP (24,25). SEs for the hospitalization estimates were calculated by using SUDAAN software and discharge weights provided by HCUP to account for the sampling design; SEs were used to calculate 95% confidence intervals (CIs) for the rates (25). If the relative SE of national estimates exceeded 0.30, or if the number of unweighted outpatient visits or hospitalizations in a strata was <30, the estimates were considered unreliable and were not shown here (26).

A weighted least-squares technique was used to assess a linear trend in the annual hospitalization rate for ulcer types during the 8-year study period of 1998–2005; the independent variable was year and the dependent variable was the rate. A modification to the classical regression technique was necessary to account for the changing NIS survey design over the study period, as described by Gillum et al. (27). In this method, a regression line is fit to the data, and the resulting slope is tested for difference from zero by using a Wald test for significance; autocorrelation was not assessed. A p value <0.05 was considered significant in this study.

Age-adjusted rates were calculated by using the direct method, taking into account the survey design, and using the projected 2000 US Census population as the reference population (28). In the direct method, the age-adjusted rate represents what the rate would be if the study population had the same age distribution as a reference population (i.e., the projected US Census 2000 age distribution). This method is used to remove confounding by age when rates are compared over time or across populations with different age distributions. To calculate age-adjusted rates, the age group–specific rates (hospitalizations/population) were multiplied by weights representing the proportion of the reference population belonging in the corresponding age group; the resulting quantities were summed to obtain the age-adjusted rate (28). We report the age-adjusted PUD hospitalization rates; the overall age-adjusted rate did not differ from the unadjusted rate, although the rates for some groups did differ.

Our analysis of hospital discharge records from a nationally representative sample of US hospitals indicates that the overall age-adjusted rate of hospitalization for PUD declined during 1998–2005. This finding is consistent with decreases previously observed for PUD hospitalizations in the United States in several studies from the 1970s through the 1990s (7,8,14,17,29). One study by Manuel et al. did not observe a downward trend in PUD hospitalizations during 1996–2005; however, that study included only 5 hospitals (30). We analyzed data for first-listed PUD hospitalizations to limit data to hospitalizations for care specifically for ulcer-related issues. In addition, >50% of patients in our study had an EGD with closed biopsy of ≥1 sites involving the esophagus, stomach, or duodenum (Figure 2). EGD with biopsy is a reliable technique to differentiate between PUD and other causes of abdominal pain such as gastritis (17). Thus, it appears that the data in this study are reflective of patients who were truly hospitalized primarily for PUD or its complications.

Decreases in PUD hospitalizations are likely attributable to an underlying decline in H. pylori prevalence (7,15). However, declines in PUD hospitalizations have also been attributed to changes in diagnosis coding because of improved diagnostic specificity associated with endoscopy (31). If the decrease in PUD hospitalizations was attributable to changes in diagnosis coding, the decrease in PUD hospitalizations would be inversely related to a rise in hospitalizations for gastritis/duodenitis. We found that hospitalization rates declined for gastritis/duodenitis and for PUD, which indicates that the results cannot be attributable to changes in diagnosis coding practices. Furthermore, because the hospitalization rate for all diagnoses did not change significantly from 1998 to 2005, the decline in the PUD hospitalization rate observed would not merely reflect general trends in hospitalization. The overall rate for any listed H. pylori diagnosis declined significantly during the study period, which suggests that a decrease in rates of H. pylori infections may be partially responsible for the decrease in hospitalizations for PUD.

In our study, the overall rate of hospitalizations for PUD differed according to the patient’s age, sex, race/ethnicity, and region. The highest rates of hospitalization for those with both PUD and H. pylori infection were for adults >65 years of age and decreased with each subsequent age group. This finding may result from an underlying birth cohort effect, in this case a decrease in H. pylori incidence for younger generations because of improved sanitation and fewer risk factors for transmission (18,32,33). A similar percentage change in rate of PUD hospitalizations was observed for all age groups >20 years. The comparable declines for these age groups may be partially attributable to increased use of H. pylori eradication therapy during 1998–2005, perhaps because of increased awareness among clinicians and patients of the association between H. pylori and PUD (9).

In this study, the overall rate of hospitalization for PUD in 1998 was higher for male patients than for female patients. However, by 2005 this difference had narrowed considerably because of a greater decrease in rates for male patients than for female patients. A 1985 study that examined data from the National Center for Health Statistics also recognized a trend toward comparable rates of hospitalization for both sexes (34). Our study also found differences in hospitalization rates between sexes by designated ulcer type; duodenal ulcer hospitalization rates were higher for male patients than for female patients. This finding is consistent with hospital admission data from the United Kingdom (15,16).

A study that used a national sample of US hospital discharge records noted differences in the hospitalization rate for PUD between racial/ethnic groups; blacks were more frequently hospitalized for PUD than whites in 1998 (13). Although we also found that rate of hospitalization for PUD was higher for blacks than for whites, the rate appears to be declining more rapidly for blacks than for whites. In addition, although rates were significantly lower for whites than for those in other racial/ethnic categories in 1998, by 2005 this rate difference was no longer significant because for whites, the decline apparently occurred more slowly than it did for all other racial/ethnic groups. Differences also varied by sex, as well as race/ethnicity, and suggest that hospitalizations for PUD among nonwhite men may merit further investigation. Race/ethnicity information was missing for patients in 26% of hospitalization records, possibly making comparisons between racial/ethnic groups inaccurate. A study of underreporting of race/ethnicity information in the National Hospital Discharge Survey suggests that hospitals that do not report race/ethnicity information may have a higher proportion of discharges for whites and a lower proportion of discharges for blacks than hospitals that do report race/ethnicity information (35). Our study did not examine PUD hospitalizations for American Indians and Alaska Natives because the survey’s sample size was not large enough and did not include visits to Indian Health Service or tribal facilities. However, a previous study showed that among American Indians and Alaska Natives, the prevalence of ulcer-associated conditions was high during 1996–2005, which indicates that hospitalizations for PUD among this group may warrant further study (36).

Our study showed similar trends for hospitalizations for PUD and H. pylori infection, although we noted some differences. For both PUD and H. pylori infections, the rate of hospitalization increased with age, and the age-adjusted hospitalization rate was lower for whites than for persons in any other racial/ethnic group category. In addition, the overall age-adjusted rate of both PUD and H. pylori hospitalizations was higher for male patients than for female patients. However, although the age-adjusted PUD hospitalization rate appears to be declining more rapidly among male patients than among female patients, the age-adjusted H. pylori infection hospitalization rate appears to be declining at a similar pace for female patients and male patients. The age-adjusted PUD hospitalization rate for Hispanics did not decline significantly, and a decline in the age-adjusted H. pylori hospitalization rate for this group was only borderline significant, which suggests that rates among this group may deserve special attention. However, this finding may be biased because of missing information on race/ethnicity.

Our findings in this study show a continued downward trend in the rate of hospitalizations for PUD in the United States. Differences in the rate of decline for PUD hospitalization rates between sexes, racial/ethnic groups, and regions warrant further study. The overall downward trend observed in this study does not seem to be attributable to increases in gastritis/duodenitis hospitalizations or to a decline in total hospitalizations. The decline in the PUD hospitalization rate may be attributable to a birth cohort effect with subsequent declines in H. pylori infection prevalence and increased use of successful antibiotic treatments to eradicate H. pylori infections. Other factors possibly contributed to the decline in PUD hospitalizations observed in this study, including trends in use of nonsteroidal anti-inflammatory drugs and the availability of over-the-counter H2 antagonists and proton pump inhibitors. Studies on the relationship between PUD hospitalizations and nonsteroidal anti-inflammatory drug use, the possibility of undercoding for H. pylori on hospitalization discharge records, and subpopulation analyses would help further guide recommendations and show how to focus interventions. To facilitate further declines in hospitalizations for PUD, patients and clinicians should continue to be educated about the association between H. pylori and PUD.