Process of Care and Outcomes for Elderly Patients Hospitalized With Peptic Ulcer Disease

Transcription

1 ORIGINAL CONTRIBUTION Process of Care and Outcomes for Elderly Patients Hospitalized With Peptic Ulcer Disease Results From a Quality Improvement Project Jane Brock, MD, MSPH Angela Sauaia, MD, PhD Dennis Ahnen, MD William Marine, MD, MPH William Schluter, MD, MSPH Beth R. Stevens, MS Jeanne D. Scinto, PhD, MPH Herbert Karp, MD Dale Bratzler, DO, MPH PEPTIC ULCER DISEASE (PUD) IS one of the most common disorders affecting the gastrointestinal system, with a lifetime cumulative prevalence of 10% 1 and a peak prevalence between ages 65 and 74 years. 2 The costs of PUD, including the indirect costs of lost work time and productivity, are estimated to be at least $8 billion per year in the United States. 3 In 1998, approximately 1.5% of all Medicare hospital costs were spent treating the consequences of PUD. 4 Infection with Helicobacter pylori is considered to be the cause of 95% to 100% of duodenal ulcers and 70% to 90% of gastric ulcers, and among persons without H pylori infection, nonsteroidal anti-inflammatory drug (NSAID) use is assumed to be the major cause. 2,5 In the elderly, use of NSAIDs is a contributing factor for up to 50% of ulcers and is associated with increased rates of ulcerrelated complications, such as perforation, bleeding, and death. 6-9 See also p 2023 and Patient Page. Context Since publication in 1994 of guidelines for management of peptic ulcer disease (PUD), trends in physician practice and outcomes related to guideline application have not been evaluated. Objectives To describe changes in process of care that occurred in a quality improvement program for patients hospitalized with PUD and to evaluate associations between in-hospital treatment of PUD and 1-year rehospitalization for PUD and mortality in a subset of these patients. Design, Setting, and Patients Cohort study of 4292 sequential Medicare beneficiaries hospitalized at acute care hospitals with a principal diagnosis of PUD in 5 states (Colorado, Georgia, Connecticut, Oklahoma, and Virginia) in 1995 (baseline) and 1997 (remeasurement); outcomes were evaluated for 752 patients in Colorado. Main Outcome Measures Changes in rates of screening for Helicobacter pylori infection, treatment for H pylori infection, screening for nonsteroidal anti-inflammatory drug (NSAID) use, counseling about NSAID use; outcomes included rehospitalization for PUD and all-cause mortality within 1 year of discharge in Colorado. Results Screening for H pylori infection increased significantly (12%-19% increase; P.001) in each of the 5 states. Treatment of H pylori infection increased in each state and was significantly increased for the entire group of hospitalizations examined (8% increase overall; P=.001). Despite increased screening, detection of H pylori infection was less frequent than expected in every state, (13%-24%) and did not increase in any state. Screening for and counseling about NSAIDs did not significantly increase overall or in any state. In the Colorado cohort, the proportion of patients rehospitalized was unchanged in 1995 (8.9%) and 1997 (6.8%), and 124 patients (16%) in the combined 1995 and 1997 cohorts died within 1 year. Treatment for H pylori was not associated with a reduction in rehospitalization within 1 year (adjusted odds ratio [OR], 1.24; 95% confidence interval [CI], ) or with a reduction in mortality (adjusted OR, 1.08; 95% CI, ). Counseling about NSAID use was associated with a decrease in risk of 1-year rehospitalization for PUD (adjusted OR, 0.47; 95% CI, ) and risk of all-cause mortality (adjusted OR, 0.44; 95% CI, ). Conclusions This quality improvement program for elderly patients with PUD resulted in increased screening for H pylori and increased treatment of H pylori infection but no change in counseling about NSAID use. However, with the low prevalence of H pylori detected, treatment of H pylori infection was not associated with a reduction in repeat hospitalization for PUD or subsequent mortality, whereas counseling about the risks of using NSAIDs was associated with a reduction in the risk of both outcomes. JAMA. 2001;286: In 1994, a National Institutes of Health (NIH) Consensus Development Conference published recommendations for the management of PUD that Author Affiliations are listed at the end of this article. Corresponding Author and Reprints: Jane Brock, MD, MSPH, Colorado Foundation for Medical Care, 2851 S Parker Rd, Suite 1000, Aurora, CO ( copro.jbrock@sdps.org, jbrockp@yahoo.com) American Medical Association. All rights reserved. (Reprinted) JAMA, October 24/31, 2001 Vol 286, No

2 Table 1. Quality Indicators* Helicobacter pylori screening reflect the new understanding of the role of H pylori infection. 5 The panel advised that all patients should be screened for H pylori and that infection should be eradicated whenever detected. The panel also recommended that all patients should be evaluated for use of NSAIDs and that use of these drugs should be eliminated whenever possible. Practice guidelines were subsequently published incorporating these recommendations and also endorsing empirical treatment of H pylori infection in patients with duodenal ulcers because of the strong association of infection with ulcers in this location. 10 Although the efficacy of each intervention in the PUD guidelines has been demonstrated in randomized controlled trials, the effectiveness of guideline implementation to change clinically relevant health outcomes 11,12 in an unselected patient population has not been measured. Recommendations to eradicate H pylori in PUD are based on randomized, controlled clinical trials that used endoscopic end points and, in many cases, excluded patients of advanced age 13,16,17 or with comorbidities 15,16 or NSAID use Recommendations to eliminate NSAID use are based on extensive experience with these drugs in patients with PUD, particularly in the elderly. 2,8,18,19 Percentage of hospitalizations in which patients were tested for H pylori using serology, carbon-labeled breath test, or gastric antral biopsy with urease test, culture, or histology. Treatment of H pylori Percentage of hospitalizations in which patients with a positive test result for H pylori received appropriate antibiotic treatment. Appropriate treatment was defined as the following combinations of bismuth, amoxicillin, metronidazole, tetracycline, clarithromycin, omeprazole, and ranitidine: bismuth-metronidazole-tetracycline, bismuth-metronidazoleamoxicillin, bismuth-tetracycline-clarithromycin, bismuthamoxicillin-clarithromycin, metronidazole-amoxicillin-ranitidine, clarithromycin-amoxicillin-ranitidine, amoxicillin-omeprazoleclarithromycin, metronidazole-omeprazole-clarithromycin, metronidazole-amoxicillin-omeprazole, omeprazole-amoxicillin, and omeprazole-clarithromycin. Screening for NSAID use Percentage of hospitalizations in which patients were screened for preadmission NSAID use. To be accepted as screened, use or non-use of NSAIDs needed to be specifically documented. Counseling about NSAID use Percentage of hospitalizations in which patients had a discussion with a clinician about use of NSAIDs. To be accepted as counseled, patient record needed to contain documentation that the patient was informed of the risks and benefits of NSAID use by a physician or a nurse. *NSAID indicates nonsteroidal anti-inflammatory drug. The 1994 NIH recommendations prompted quality improvement projects (QIPs) within the Health Care Quality Improvement Program for Medicare beneficiaries led by the Health Care Financing Administration (HCFA; now the Centers for Medicare and Medicaid Services). Health Care Quality Improvement Program projects are intended to improve practice by encouraging compliance with national guidelines. 20 For this report, we obtained information from a peptic ulcer disease QIP performed by 5 state peer review organizations: Colorado Foundation for Medical Care, Connecticut Peer Review Organization, Georgia Medical Care Foundation, Oklahoma Foundation for Medical Quality, and Virginia Health Quality Center. The specific objectives of the QIP were to measure and improve the practice of testing for and treating H pylori and to measure and improve the practice of screening for and counseling about the risks of NSAID use. Practice was measured in hospital cases from 1995 (baseline) and in 1997 (remeasurement). Improvement was planned through data feedback vs a focused continuing medical education program tested by a randomized controlled trial (the results of this intervention evaluation will be published elsewhere). The assessment of baseline practice patterns from 1995 has been previously published. 21 The patient population in this QIP is unselected, is older and sicker than populations used in clinical trials, and represents patients with PUD who consume the most resources, have the highest rate of poor outcomes, 2 and presumably have the most to gain from effective treatment. The purposes of this study were (1) to describe changes in management of PUD in a cohort of Medicare patients during a multistate QIP and (2) to explore associations between PUD management and rehospitalization for PUD and all-cause mortality at 1 year after discharge for the index hospitalization in the Colorado cohort. METHODS Quality-of-Care Indicators Indicators used to measure quality of care were derived from the recommendations from the NIH consensus panel (TABLE 1). Information regarding treatment of H pylori infection was collected for all patients, including those who were not screened. A detailed description of this process has been published previously. 21 Sample Selection All records were identified by searching the Medicare claims files for hospital discharges with principal International Classification of Diseases, 9th Edition, Clinical Modification (ICD-9- CM) codes through The Medicare claims files contain information about Medicare enrollees who are not enrolled in managed care plans. At the time of the study, 75% of all Medicare beneficiaries in Colorado and more than 95% of Medicare beneficiaries in the other 4 states were insured through the fee-for-service system. 23 For the initial assessment of practice, each peer review organization started with a discharge date of June 30, 1995, and sampled in reverse chronological order until 550 records were accumulated, which was the number required according to power calculations. Selection criteria were patient age of at 1986 JAMA, October 24/31, 2001 Vol 286, No. 16 (Reprinted) 2001 American Medical Association. All rights reserved.

3 least 65 years, principal hospitalization diagnosis of PUD, and discharge alive from an acute care facility. Hospitalizations were excluded if the patient died during the hospitalization, left the hospital against medical advice, or was transferred to another care facility. For repeat measurement of practice, each peer review organization started with a discharge date of June 30, 1997, and again searched in reverse chronological order for the number of records needed to meet power requirements. Ninety-six percent of all requested records were received for abstraction. Because hospitalizations were the sampling unit, the sample contained more than 1 hospitalization for some patients. Data Collection Data were obtained from 2 sources: medical record abstraction and administrative claims data. Medical Record Abstraction. Data collected from medical records were used to evaluate practice patterns. Screening tests for H pylori included serology, culture, histological or urease testing of biopsy specimens, or carbonlabeled urea breath testing. Helicobacter pylori infection was recorded in 1 of 3 ways: as positive if any test result was positive, as negative if all test results were negative, or as not documented if the result of any given test was not found in the record. Ulcer-related risk factors that were abstracted included human immunodeficiency virus infection, acquired immunodeficiency syndrome, hepatic failure, leukemia, metastatic cancer, gastrointestinal cancer, immunosuppression, cirrhosis, and Zollinger-Ellison syndrome. Screening for NSAID use was considered to be positive if an NSAID was listed by the physician as a preadmission medication or specific documentation existed referring to the use or nonuse of NSAIDs. If that documentation referred to recent or current use, that patient was considered to be a current user of the medication. Performance of screening for NSAID use was abstracted from both the 1995 and 1997 records, but recent/current use of NSAIDs was abstracted only from the 1995 records. Counseling regarding NSAID use was considered to be positive if the record contained documentation that the patient was informed of the risks and benefits of NSAID use by either a physician or nurse. (All references to screening and treatment throughout the remainder of this article indicate documentation of screening and treatment.) Location of ulcer was abstracted from the 1997 charts based on the 1996 guidelines that indicated empirical treatment for H pylori in patients with duodenal ulcers was acceptable. 10 Administrative Claims Data. Data from administrative claims records were used for analyses of rehospitalization for PUD and mortality in the Colorado QIP cohort. Secondary diagnosis and procedure codes assigned during the index hospitalization were available for Colorado beneficiaries from the Colorado Medicare claims files. Medicare files include dates of both beneficiary enrollment in a managed care plan and beneficiary death. Outcomes of subsequent hospitalization and death within 1 year were obtained by searching the Medicare files for these events. To link patient outcomes with treatment, data collected by hospitalizations were attributed to individual patients because the 846 abstracted hospitalizations in Colorado included 30 second hospitalizations (4%) for patients already included in the sample (FIGURE). The data in the QIP therefore represented the treatment experiences of 816 patients. Of those, 11 were lost from the claims files and 6 had incomplete claims information. No additional records were removed from the analysis of death within 1 year. All patients who died or joined a managed care plan within 1 year without a repeat admission for PUD (n=173) were excluded from the analysis of rehospitalization for PUD because data regarding their rehospitalizations would not be available in the Colorado Medicare files. Ulcer location, based on ICD-9-CM codes, was gastric in 51% (n=410) of the cases and duodenal in 43% Figure. Selection of Cases in Colorado Cohort for Evaluation of Readmission and Mortality 894 Hospitalization Claims With Diagnosis of PUD 48 Exclusions 12 Not PUD 36 Other 846 Hospitalizations Abstracted 30 Records Excluded (Second Hospitalizations for Patients Already in Sample) 816 First Hospitalizations Abstracted 799 With Subsequent Information Available in the Claims File 752 With Gastric or Duodenal Ulcers 626 Hospitalizations With 1 y of Follow-up Information in the Claims File 595 With Gastric or Duodenal Ulcers 48 Readmitted Within 1 y of Discharge 17 Records Excluded 11 No Further Claims After First Admission Abstracted 6 With Incomplete Claims Data 173 Exclusions 121 Died Within 1 y Without Rehospitalization for PUD 48 Joined Managed Care Plans Within 1 y Without Rehospitalization for PUD 4 Joined Managed Care Plans and Died Within 1 y Without Rehospitalization for PUD PUD indicates peptic ulcer disease. (n=342). Location was coded as gastrojejunal in 2% (n=13) and site unspecified in 4% (n=34). Because comparing duodenal to gastric ulcers was of clinical interest, outcomes analysis was limited to the 752 patients with either gastric or duodenal ulcers. Comorbidities associated with mortality were scored from secondary ICD- 9-CM codes included in the initial abstracted hospitalization according to the method of Charlson et al and Deyo et al. 24,25 Briefly, a score of 1 was assigned for myocardial infarction, con American Medical Association. All rights reserved. (Reprinted) JAMA, October 24/31, 2001 Vol 286, No

4 gestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, connective tissue disease, mild liver disease, or diabetes; a score of 2 was assigned for hemiplegia, moderate or severe renal disease, diabetes with end organ damage, any solid tumor, leukemia, or lymphoma; a score of 3 was assigned for moderate or severe liver disease and metastatic solid tumor; and a score of 6 was assigned for acquired immunodeficiency syndrome. The sum of the patient scores was used in the analysis of death within 1 year of discharge. The comorbidity index was designed specifically to assess risk of mortality and, therefore, was not used in the analysis of risk for rehospitalization. Reliability Interstate reliability was evaluated by each state submitting 7 records that were copied and redistributed to each state for reabstraction by the lead abstractor after project abstraction was completed. values for the assessment of H pylori screening, H pylori treatment, and screening for NSAID use were greater than 0.75, indicating excellent agreement. 26 The value for the assessment of NSAID counseling was 0.64, indicating fair-to-good agreement. 26 Interrater reliability in Colorado was assessed by 2 blinded Colorado Foundation for Medical Care abstractors who scored 70 records that were then randomly redistributed through the sample for routine abstraction. Values for both H pylori treatment and NSAID counseling were 1. Duodenal ulcer location by ICD-9-CM codes showed 84% agreement with duodenal location by abstraction in the 1997 records. Ulcer location as determined by ICD-9-CM codes was therefore used in the analysis of outcomes. Sample Size and Power Calculations For the multistate project, the state sample sizes (state ranges, at baseline and at remeasurement) were calculated to detect a 10% change in performance of the quality indicators with 80% power and 95% confidence. For the outcome of readmission in the Colorado cohort, the study had 95% power to detect a difference of 10% or more between treated and untreated patients, assuming that untreated patients have a 5% risk of ulcer rebleeding. This is smaller than the expected difference of 30% between treated and untreated patients based on literature review and is a conservative estimate of the effect of H pylori treatment, which should be associated with a 0% to 5% recurrence For the outcome of death in the Colorado cohort, the study had a 90% power to detect a change of at least 10% associated with H pylori treatment, assuming a 16% risk of one-year mortality in the treated population. Statistical Analyses All analyses were performed using SAS version 6.12 (SAS Institute Inc, Cary, NC). Comparisons of quality indicator performance per period within states were examined using 2 tests, and the Fisher exact test was used for comparisons in categorical variables with small cell counts. Age was analyzed as a continuous variable using analysis of variance, and pairwise comparisons of means were tested with significant differences determined at P.05. Multiple logistic regression was used to examine the independent associations between treatment of H pylori, and counseling about NSAID use on the outcomes of interest. Any variable for which the univariate test had a P.25 was considered to be a possible confounder and, therefore, was a candidate for the multivariate model based on the methods of Hosmer and Lemeshow. 31 Analysis of mortality was adjusted for age greater than 74 years, presence of duodenal ulcer, presence of perforation, and Charlson comorbidity score. The independent associations of the 2 treatment variables, H pylori treatment and counseling for NSAID use, with the 2 outcomes of interest were examined in the entire Colorado cohort with gastric or duodenal ulcers, the subset of patients who had a positive test for H pylori infection, and the subset of patients who were identified as recent NSAID users. RESULTS Multistate Project Description of Hospitalizations. Abstraction was completed for 2644 hospitalizations during 1995 and 1648 hospitalizations in The mean (SD) age of the patient samples was between 76.6 (7.7) and 78.6 (7.5). Both samples from Georgia were younger than the other state samples (76.6 [7.6] and 76.7 [7.3] years compared with 77.4 [7.8] and 78.6 [7.5] years; P.05). Proportions of women were similar among the samples (50% and 64%); frequency of nonwhite persons had greater variation (5%-20%), as would be expected from the demographics of the participating states. Ulcer-related risk factors were present in 8% to 14% of abstracted samples. Screening and Treatment for PUD Risk Factors. Rates of screening for H pylori increased between 1995 and 1997 in all states (12%-19% increase; P=.001 for each; TABLE 2). Treatment of H pylori infection during hospitalizations for PUD increased in all states and increased significantly in Colorado (9% increase; P =.002) and Oklahoma (13% increase; P=.001). Helicobacter pylori infection was detected in 13% to 24% of hospitalizations and did not change significantly from 1995 to 1997 in any state. Although most states showed increases in rates of treatment for patients with H pylori infection, the increases were not statistically significant, as rates of treating infected patients were relatively high at baseline. Recent NSAID use was identified in 57% to 68% of patients in the 1995 sample. There were no significant increases among the states in rates of screening for or counseling about NSAID use. Colorado Cohort Screening and Treatment for PUD Risk Factors. Of the Colorado sample of 799 patients, 60% (n=479) were screened 1988 JAMA, October 24/31, 2001 Vol 286, No. 16 (Reprinted) 2001 American Medical Association. All rights reserved.

5 Table 2. Performance of Screening and Treatment of Peptic Ulcer Disease Risk Factors by State and Time Interval* Quality Indicators, % Colorado Connecticut Georgia Oklahoma Virginia Baseline Follow-up P Baseline (n = 502) (n = 341) Value (n = 531) Follow-up (n = 533) P Baseline Value (n = 533) Follow-up (n = 318) P Baseline Value (n = 549) Follow-up (n = 354) P Baseline Value (n = 529) Follow-up (n = 249) Helicobacter pylori infection Screened Treated Confirmed Percentage of infected treated NSAID use Screened Recent Counseled *P values refer to the comparison between baseline and follow-up period performance for each state. Baseline data collection occurred in 1995 and follow-up data collection occurred in NSAID indicates nonsteroidal anti-inflammatory drug. NSAID use status determined during baseline interval (1995) only. P Value for H pylori infection and 26% (n=208) were treated for H pylori infection. Treatment of H pylori was not always associated with screening; of the 208 patients treated, 74% (n=154) had been screened and 26% (n=54) had been treated without a screening test. Among patients with diagnosed infection (n=121), 78% (n=94) were treated for H pylori or had plans for treatment recorded in the chart. Eightytwo percent (n=657) were screened for NSAID use and 29% (n=228) were counseled about use of NSAIDs. Among patients in the 1995 cohort who were identified as recent users of NSAIDs (n=319), 37% (n=119) were counseled about the associated risks. Patients With Gastric or Duodenal Ulcers. Of 799 patients in the Colorado cohort, 752 (94%) had ulcer location of gastric or duodenal recorded in the chart. The mean age of these 752 patients was 78 years (range, years) and 56 % were female (TABLE 3). Fiftyone percent (n=386) had a total Charlson comorbidity score of 0 and 13% (n=97) had at least 1 ulcer-related risk factor. Bleeding was much more common than perforation (78% and 8%). Helicobacter pylori infection was diagnosed in 16% (n=118) of the entire group, or 26% of those who were tested. Fifty-four percent (n=251) of those tested had a negative test result for H pylori, and results were not documented in the chart for 20% (n=93). Recent use of Table 3. Characteristics of 752 Patients With Gastric or Duodenal Ulcer in Colorado Cohort Total (n = 752) NSAIDs was documented in 82% (n=297) of the 362 patients for whom that variable was abstracted. Duodenal ulcers were not significantly more likely to be associated with H pylori infection than gastric ulcers (29% vs 23%; P=.20). Outcomes in the Colorado Cohort Readmission for PUD Within 1 Year. Forty-eight patients (8% of those with documented gastric or duodenal ulcer; n=595) in the Colorado cohort were readmitted for PUD within one year of discharge. The proportion of patients readmitted at 1 year was not significantly different between the baseline and remeasurement samples (8.9% Gastric Ulcer (n = 410) Duodenal Ulcer (n = 342) P Value Age, mean (SD), y 78.5 (7.7) 78.8 (7.4) 78.0 (7.9).22 Female, % Charlson comorbidity score 0, %* Ulcer-related risk factor, % Bleeding, %* Perforation, %* Helicobacter pylori infection, % of tested (n = 462) Positive Negative Not documented Recent nonsteroidal anti-inflammatory drug use, % of screened (n = 362) Positive Negative *Charlson comorbidity score, ulcer location, bleeding, and perforation were determined from International Classification of Diseases, Ninth Revision, Clinical Modification codes. Determined from 1995 data only. in 1995 vs 6.8% in 1997; P=.36). Univariate analysis of treatment variables, demographic characteristics, and ulcerrelated risk factors were not significantly associated with repeat admission for PUD (TABLE 4). Univariate analysis of H pylori treatment in patients with known infection (n=92) and NSAID counseling in NSAID users (n=243) showed no significant associations with readmission rates. In univariate analysis, ulcer location was the only variable with a P 0.25 and was therefore entered as a potential confounder in the model. In the regression model adjusting for ulcer location, treatment for H pylori was not 2001 American Medical Association. All rights reserved. (Reprinted) JAMA, October 24/31, 2001 Vol 286, No

6 associated with reduced risk of rehospitalization for PUD (adjusted odds ratio [OR], 1.24; 95% confidence interval [CI], ; P=.52), whereas counseling about NSAID use was associated with a significant reduction in risk of readmission for PUD (OR, 0.47; 95% CI, ; P=.048). There was no association between H pylori treatment and risk of rehospitalization for PUD in patients with H pylori infection (OR, 0.96; 95% CI, ; P=.97) or between NSAID counseling in patients documented to be recent users of NSAIDs (OR, 0.62; 95% CI, ; P=.38). Mortality Within One Year. Combining 1995 and 1997 data, 124 patients (16%) in the Colorado cohort died within 1 year of discharge (Table 4). Univariate analysis showed no association between treatment of H pylori infection and mortality (16% of treated vs 17% of untreated; P=.88). One-year mortality was significantly less in patients who were counseled about NSAID use (9% vs 20%; P=.001) and Table 4. Proportion of Patients Who Died or Were Readmitted Within 1 Year, by Selected Case Characteristics* Death Readmission Total Died Within Unadjusted Total Readmitted Within Characteristics Patients, No. 1 Year, No. (%) P Value Patients, No. 1 Year, No. (%) Overall (16) (8) Treatment Treated for Helicobacter pylori (16) (9).88 Not treated for H pylori (17) (8) Counseling Counseled about NSAID use (9) (5).001 Not counseled about NSAID use (20) (10) Age, y (18) (8) (13) (7) Unadjusted P Value Female (16) (7).51 Charlson comorbidity score (10) (6) (20) (11) (24) 67 7 (10) (35) 32 2 (6) Ulcer-related risk factors Yes (27) 71 5 (7).003 No (15) (8).74 Location of ulcer Duodenal (20) (10).02 Gastric (14) (7).15 Bleeding Yes (17) (7).25 No (14) (10).32 Perforation Yes (27) 41 3 (7).02 No (16) (8).90 H pylori infection in those screened Positive (16).62# 92 9 (10).41# Treated (17) 70 7 (10).47 Not treated 26 3 (12) 22 2 (9) 1.0 Negative (17).18# (7).80# NSAID use in those for whom use status was recorded Recent user (14) (7) Counseled (6) 99 5 (5) Not counseled (19) (8) Not recent user 65 4 (6) 57 4 (7) *NSAID indicates nonsteroidal anti-inflammatory drug. Analysis is based on all 752 patients in the Colorado cohort. P values were determined using 2 test. Analysis based on 595 patients in the Colorado cohort with gastric or duodenal ulcers (see Figure). P value for trend. Ulcer location, bleeding, and perforation were determined from International Classification of Diseases, Ninth Revision, Clinical Modification codes. #For these P values, positive results were compared with negative results plus no results, and negative results were compared with positive results plus no results JAMA, October 24/31, 2001 Vol 286, No. 16 (Reprinted) 2001 American Medical Association. All rights reserved.

7 significantly more in patients with a Charlson comorbidity score greater than 0 (23% vs 10% with a score of 0; P=.001), ulcer-related risk factors (27% vs 15% without; P=.003), duodenal ulcers (20% vs 14% without; P=.02), and perforation (27% vs 16% without; P=.02). Univariate analysis in the subset of patients with proven infection (n=118) also showed no significant difference in mortality based on H pylori treatment (17% of treated vs 12% of untreated; P=.47). Univariate analysis of counseling about NSAIDs in the subset of patients who were NSAID users (n=297) was associated with a significant reduction in mortality (6% for those counseled vs 19% for those not counseled; P=.001). Multiple logistic regression analysis (adjusted for age 74 years, Charlson comorbidity score 0, presence of ulcer-related risk factors, duodenal ulcer location, and presence of perforation) showed no reduction in mortality risk associated with treatment of H pylori (adjusted OR, 1.08; 95% CI, ; P=.74), but a significant reduction in mortality risk associated with counseling about NSAID use (adjusted OR, 0.44; 95% CI, ; P=.003). When the same model was repeated for the subset of patients with H pylori infection (n=118), there was no difference in mortality risk based on treatment (adjusted OR, 1.65; 95% CI, ; P=0.48). When modeling was repeated in the subset of patients who were recent NSAID users counseling about NSAID use was associated with a reduction in mortality risk (adjusted OR, 0.3; 95% CI, ; P =.01). Independent predictors of increased risk of mortality included age greater than 74 years (OR, 1.59; 95% CI, ), Charlson comorbidity score greater than 0 (OR, 2.31; 95% CI, ), ulcer-related risk factors (OR, 1.80; 95% CI, ), and duodenal ulcer location (OR, 1.52; 95% CI, ). COMMENT In this QIP for treatment of patients with PUD, rates of screening for and treatment of H pylori infection increased in all 5 states between 1995 and Detection of H pylori did not change as a result of improved screening in any state, and screening and counseling for NSAID use did not increase in any state. Guidelines for treatment of PUD strongly recommend screening and counseling all patients about the risks of NSAID use. 10,32 Although adoption of guidelines is a complex and incompletely understood process, 33,34 the impetus for publishing guidelines was the new knowledge of the role of H pylori infection, giving treatment for infection greater emphasis. This knowledge was based on extensive literature demonstrating that H pylori infection is the major cause of PUD; 1 such article stated that treatment for H pylori was the cure. 14 The rate of screening and treating H pylori infection did increase during the QIP period, suggesting that dissemination of the main message was effective. The randomized controlled trials that established the role of H pylori in PUD were performed in populations that were not representative of the population included in this QIP. The guidelines had not been tested in unselected populations prior to publication. Both of these potential limitations of guidelines are better appreciated today than in 1994, when the NIH guidelines were published and this QIP was designed. 12 The population in this QIP did not have the expected prevalence of H pylori, and counseling about NSAID use did not increase as a result of the project. In retrospect, this QIP could have included concurrent sample data analysis to search for unanticipated results, as is more typical of present interventions to improve quality In the analysis of the Colorado cohort, treatment of H pylori was not associated with a reduction in either 1-year mortality or readmission for PUD in 1 year in the population. Despite the relatively small number of patients with diagnosed infection (n=92), these data have an 80% power to detect a 22% difference in readmission rates due to H pylori treatment (2-tailed =.05); a 22% reduction is less than the 27% to 33% reduction in recurrence of bleeding usually attributed to H pylori eradication in the infected population. Counseling about the use of NSAIDs was associated with a significant reduction in 1-year readmission for PUD. The analysis of mortality supports the results from the readmission analysis, which is that lack of counseling about NSAID use is more strongly associated with poor outcomes in this patient population than is treatment of H pylori infection. The impetus for assessing mortality was to determine if adverse effects of treatment for PUD were occurring in an elderly population in which PUD might not be the primary determinant of mortality. However, these data are clearly unsuited for a definitive assessment of all-cause mortality; administrative data lack information on cause of death and medications received for PUD or other illnesses, and data were not collected for analysis of mortality. It is possible that counseling about NSAID use is a proxy for other therapies to reduce NSAID-mediated injury, such as maintenance acid suppression, 38 proton pump inhibitors, 39 or misoprostol. 40 Documentation of counseling about NSAIDs most likely was related to general quality of care, with patients who received this intervention receiving better care in general. If true, this emphasizes that quality care for these patients included counseling about NSAID use and was more related to improved outcomes than was treatment for H pylori infection. There was a relatively low frequency of H pylori infection detected in this patient population (26% among those screened), which is lower than documented rates in previous studies. 2,3,13-17 This low rate of H pylori infection in hospitalized elderly patients with PUD was seen in all states, ranging from 15% (27% of all tested) in Oklahoma to 24% (34% of all tested) in Virginia. Because the data in our study are from hospitalizations, some patients may have been previously treated for H pylori infection as outpa American Medical Association. All rights reserved. (Reprinted) JAMA, October 24/31, 2001 Vol 286, No

8 tients. If they were, however, it did not prevent their need for hospital admission for PUD. Previous work has shown a lower prevalence of H pylori infection in complicated PUD than in uncomplicated PUD, 41 but rates in this QIP study population are much lower than reported in previous studies of complicated PUD. Hosking et al 42 reported a prevalence of 71% in a population of hospitalized patients with bleeding duodenal ulcers Jensen et al 43 reported 74% of patients admitted with upper gastrointestinal tract bleeding due to ulcers were infected. We estimate that every patient in the Colorado cohort who was not tested for H pylori plus every patient with an undocumented result of H pylori testing would have had to be infected to equal the prevalence anticipated based on previous research. Moreover, rates of readmission and death in H pylori infected patients who were not treated (9% were readmitted and 12% died) were not increased, suggesting little room for improvement in outcomes due to treatment of H pylori infection. An alternative explanation for our results is that physicians may counsel patients with more favorable prognoses, thereby associating counseling with better outcomes. It is unlikely to be the entire explanation, however, because the associations between counseling about NSAID use and outcomes were essentially unchanged after adjusting for other prognostic factors. This study has the limitation of using data obtained by chart abstraction in that all determination of screening and treatment depends on documentation of those interventions. These data do not appear to have a survivor bias, despite the exclusion of those who died in the hospital, were transferred to another acute care facility, or left against medical advice (n=36 hospitalizations of 894 hospital records originally received for abstraction). A sensitivity analysis assuming that all 36 of these hospitalizations represented persons hospitalized with H pylori infection who were not treated and died does not substantially change our results that treatment for H pylori was not associated with improved survival at 1 year. In summary, this QIP for elderly patients with PUD led to increased screening for and treatment of H pylori infection. However, treating H pylori did not influence outcomes for this elderly population hospitalized with PUD, most likely because H pylori infection was not the main factor causing ulcers in these elderly patients. This study suggests that counseling about NSAID use in similar patients may reduce 1-year readmission for PUD. Future projects to improve the management of elderly patients hospitalized with PUD should focus on increasing physician awareness of the role of NSAIDs in PUD and improving rates of educating and treating patients about the associated risks. Furthermore, future projects involving guideline implementation should carefully consider the target population, and guidelines should state clearly to what populations they may be applicable and whether they have been tested in unselected populations. As with other clinical interventions, guideline implementation should involve concurrent data analysis with the capability to change the intervention in the case of unexpected results. Author Affiliations: Colorado Foundation for Medical Care, Aurora (Drs Brock, Sauaia, and Schluter and Ms Stevens); Department of Preventive Medicine, University of Colorado School of Medicine, Denver (Drs Brock, Sauaia, Ahnen, Marine, and Schluter); Department of Gastroenterology, Denver Veteran s Administration Hospital, Denver, Colo (Dr Ahnen); Qualidigm, Middletown, Conn (Dr Scinto); University of Connecticut Health Center School of Medicine, Farmington (Dr Scinto); Georgia Medical Care Foundation, Atlanta (Dr Karp); and Oklahoma Foundation for Medical Quality, Oklahoma City (Dr Bratzler). Author Contributions: Study concept and design: Brock, Sauaia, Ahnen, Schluter. Acquisition of data: Stevens, Karp, Bratzler. Analysis and interpretation of data: Brock, Sauaia, Ahnen, Marine, Stevens, Scinto. Drafting of the manuscript: Brock, Sauaia. Critical revision of the manuscript for important intellectual content: Brock, Sauaia, Ahnen, Marine, Schluter, Stevens, Scinto, Karp, Braztler. Statistical expertise: Brock, Sauaia, Stevens, Scinto. Administrative, technical, or material support: Marine, Schluter, Scinto, Bratzler. Study supervision: Sauaia, Ahnen, Marine, Schluter, Scinto. Funding/Support: Support for this research was provided in part by Health Resources and Services Administration/Preventive Medicine Training grant 5D33HP Disclaimer: The analyses on which this article are based were performed under contract P11, entitled Utilization and Quality Control Peer Review Organization for the State of Colorado, sponsored by the Health Care Financing Administration, Department of Health and Human 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 endorsement by the US government. The authors assume 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 identification of QIPs derived from analysis of patterns of care, and therefore required no special funding on the part of this contractor. Ideas and contributions to the authors concerning experience in engaging with issues presented are welcomed. Acknowledgment: We acknowledge those who were fundamental to the development and completion of the quality improvement project: Michelle Mills and the staff of the Colorado Foundation for Medical Care, Aurora; Donna Piatt, RN, Wato Nsa, MD, PhD, and the staff of the Oklahoma Foundation for Medical Quality, Oklahoma City; and the staff members of the Georgia Medical Care Foundation, Atlanta, the Connecticut Peer Review Organization, Middletown, and the Virginia Health Quality Center, Richmond. REFERENCES 1. Sonnenberg A Everhart JE. The prevalence of selfreported peptic ulcer in the United States. Am J Public Health. 1996;86: Sonnenberg A. Peptic ulcer. In: Everhart JE, ed. Digestive Diseases in the United States: Epidemiology and Impact. Washington, DC: US Dept of Health and Human Services; 1994: Publication NIH Brown DM, Everhart JE. Cost of digestive diseases in the United States. In: Everhart JE, ed. Digestive Diseases in the United States: Epidemiology and Impact. Washington, DC: US Dept of Health and Human Services; 1994: Publication NIH Medicare Provider Analysis and Review (MEDPAR) of Short-Stay Hospitals. Available at: Accessed April NIH Consensus Development Panel on Helicobacter pylori in Peptic Ulcer Disease. Helicobacter pylori in peptic ulcer disease. JAMA. 1994;272: Graham DY, Lidsky MD, Cox AM, et al. Longterm nonsteroidal anti-inflammatory drug use and Helicobacter pylori infection. Gastroenterology. 1991; 100: Loeb DS, Talley NJ, Ahlquist DA, Carpenter HA, Zinsmeister AR. Long-term nonsteroidal antiinflammatory drug use and gastroduodenal injury: the role of Helicobacter pylori. Gastroenterology. 1992; 102: Greene JM, Winickoff N. Cost-conscious prescribing of nonsteroidal anti-inflammatory drugs for adults with arthritis. Arch Intern Med. 1992;152: Laine L. Acute and chronic gastrointestinal bleeding. In: Sleisenger MH, Fordtran JS, Scharschmidt BF, eds. Sleisenger and Fordtran s Gastrointestinal and Liver Disease. 6th ed. Philadelphia, Pa: WB Saunders; 1998: Soll AH, for the Practice Parameters Committee of the American College of Gastroenterology. Medical treatment of peptic ulcer disease. JAMA. 1996; 275: Haddix A, Teutsch S. Prevention Effectiveness A Guide to Decision Analysis and Economic Evaluation. New York, NY: Oxford University Press; JAMA, October 24/31, 2001 Vol 286, No. 16 (Reprinted) 2001 American Medical Association. All rights reserved.

9 12. Wilson MC, Hayward RS, Tunis SR, Bass EB, Guyatt G. Users guides to the medical literature, VIII: how to use clinical practice guidelines, B: what are the recommendations and will they help you in caring for your patients? JAMA. 1995;274: Coughlan JG, Gilligan D, Humphries H, et al. Campylobacter pylori and recurrence of duodenal ulcers: a 12-month follow-up study. Lancet. 1987;2: Rauws EA, Tytgat GN. Cure of duodenal ulcer associated with eradication of Helicobacter pylori. Lancet. 1990;335: Hentschel E, Brandstatter G, Dragosics B, et al. Effect of ranitidine and amoxicillin plus metronidazole on the eradication of Helicobacter pylori and the recurrence of duodenal ulcer. N Engl J Med. 1993; 328: Marshall BJ, Goodwin CS, Warren JR, et al. Prospective double-blind trial of duodenal ulcer relapse after eradication of Campylobacter pylori. Lancet. 1988;2: Graham DY, Lew GM, Klein PD, et al. Effect of treatment of Helicobacter pylori infection on the longterm recurrence of gastric or duodenal ulcer. Ann Intern Med. 1992;116: Paulus HE. FDA Arthritis Advisory Committee meeting: serious gastrointestinal toxicity of nonsteroidal anti-inflammatory drugs. Arthritis Rheum. 1998; 31: Gabriel SE, Jaakkimainen L, Bombardier C. Risk for serious gastrointestinal complications related to use of nonsteroidal anti-inflammatory drugs. Ann Intern Med. 1991;115: Jencks SF, Wilensky GR. The Health Care Quality Improvement Initiative. a new approach to quality assurance in medicine. JAMA. 1992;268: Ofman JJ, Etchason J, Alexander W, et al. The quality of care for medicare patients with peptic ulcer disease. Am J Gastroenterol. 2000;95: International Classification of Diseases, Ninth Revision, Clinical Modification. Washington, DC: US Dept of Health and Human Services; Dicker RC, Han LF, Macone JJ. Quality of Care Surveillance Using Administrative Data, Baltimore, Md: Health Care Financing Administration; Quality resume Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40: Deyo RA, Charkin DC, Ciol MA. Adapting a clinical comorbidity index for use with IDC-9-CM administrative databases. J Clin Epidemiol. 1992;45: Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics. 1997;33: Graham DY, Hepps KS, Ramirez FC, Lew GM, Saeed ZA. Treatment of H pylori reduces the rate of rebleeding in peptic ulcer disease. Scand J Gastroenterol. 1993;28: Jaspersen D, Koerner T, Schorr W, Brennenstuhl M, Raschka C, Hammar CH. Helicobacter pylori eradication reduces the rate of rebleeding in ulcer hemorrhage. Gastrointest Endosc. 1995;41: Labenz J, Borsch G. Role of Helicobacter pylori eradication in the prevention of peptic ulcer bleeding relapse. Digestion. 1994;55: Rokkas T, Karameris A, Mavrogeorgis A, Rallis E, Giannikos N. Eradication of Helicobacter pylori reduces the possibility of rebleeding in peptic ulcer disease. Gastrointestinal Endosc. 1995;41: Hosmer DW, Lemeshow S. Applied Logistic Regression. New York, NY: John Wiley & Sons; 1989: The European Helicobacter pylori Study Group. Current European concepts in the management of Helicobacter pylori infection: the Maastricht Consensus Report. Gut. 1997;41: Bero LA, Grilli R, Grimshaw JM, Harvey E, Oxman AD, Thomson MA. Closing the gap between research and practice: an overview of systematic reviews of interventions to promote the implementation of research findings. BMJ. 1998;317: Smith WR. Evidence for the effectiveness of techniques to change physician behavior. Chest. 2000; 118(suppl):8S-17S. 35. Berwick DM. Developing and testing changes in delivery of care. Ann Intern Med. 1998;128: Pillars of quality the Peer Review Organization/ Quality Improvement Organization Program. Available at: Accessed September 30, Elements of interventions. Available at: Accessed September 30, Davies J, Collins AJ, Dixon SA. The influence of cimetidine on peptic ulcer in patients with arthritis taking anti-inflammatory drugs. Br J Rheumatol. 1986; 25: Yeomans ND, Tulassay Z, Juhasz L, et al. A comparison of omeprazole with ranitidine for ulcers associated with nonsteroidal anti-inflammatory drugs. N Engl J Med. 1998;338: Silverstein FE, Graham DY, Senior JR, et al. Misoprostol reduces serious gastrointestinal complications in patients with rheumatoid arthritis receiving nonsteroidal anti-inflammatory drugs. Ann Intern Med. 1995;123: Lanas AI, Remacha B, Esteva F, Sainz R. Risk factors associated with refractory peptic ulcers. Gastroenterology. 1995;109: Hosking SW, Yung MY, Chung SC, Li AK. Differing prevalence of Helicobacter pylori in bleeding and nonbleeding ulcers [abstract]. Gastroenterology. 1992;102:A Jensen DM, You S, Pelayo E. The prevalence of Helicobacter pylori and nonsteroidal antiinflammatory drug use in patients with severe upper gastrointestinal hemorrhage and their potential role in recurrence of ulcer bleeding [abstract]. Gastroenterology. 1992;102:A90. Freedom! A fine word when rightly understood. What freedom would you have? What is the freedom of the most free? To act rightly! Johann Wolfgang von Goethe ( ) 2001 American Medical Association. All rights reserved. (Reprinted) JAMA, October 24/31, 2001 Vol 286, No