| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
(Circulation. 2007;116:2280-2287.)
© 2007 American Heart Association, Inc.
Health Services and Outcomes Research |
From Health Services Research and Development Center of Excellence, Ann Arbor VA Medical Center, and Department of Internal Medicine, University of Michigan Medical School (B.K.N.), Ann Arbor, Mich; Department of Medicine (Y.W., H.M.K.) and Section of Health Policy and Administration (H.M.K.), Department of Epidemiology and Public Health, Yale University School of Medicine, New Haven, Conn; Department of Medicine, University of Iowa (P.C.), Iowa City, Iowa; Department of Surgery, University of Michigan (J.D.B.), Ann Arbor, Mich; Department of Geriatrics and Adult Development, Mount Sinai School of Medicine, New York, NY and HSR&D Targeted Research Enhancement Program and the Geriatrics Research, Education, and Clinical Center, James J. Peters VA Medical Center, Bronx, NY (J.S.R.); Department of Health Care Policy, Harvard Medical School and Department of Biostatistics, Harvard School of Public Health (S.-L.T.N.), Boston, Mass; and Robert Wood Johnson Clinical Scholars Program, Yale University School of Medicine and the Yale-New Haven Hospital Center for Outcomes Research and Evaluation (H.M.K.), New Haven, Conn.
Correspondence to Dr Nallamothu, 1500 E Medical Center Dr, Ann Arbor, MI 48109-0366. E-mail bnallamo{at}umich.edu
Received April 20, 2007; accepted August 28, 2007.
| Abstract |
|---|
|
|
|---|
Methods and Results— From 2003 Medicare data, we used hierarchical regression to calculate 30-day standardized mortality ratios and risk-standardized mortality rates for AMI and CHF at 16 cardiac and 121 peer general hospitals in 15 healthcare markets. We then compared cardiac and general hospitals by determining (1) the proportion of facilities with statistically higher, no different, or lower than expected mortality based on 95% interval estimates of standardized mortality ratios and (2) differences in risk-standardized mortality rates between the types of facilities after stratification within healthcare markets. We identified 1912 patients with AMI and 1275 patients with CHF at cardiac hospitals and 13 158 patients with AMI and 18 295 patients with CHF at general hospitals. Patients at cardiac hospitals were younger, were more likely to be male, and had a much lower prevalence of noncardiovascular diseases. After adjustment for patient differences, standardized mortality ratios were significantly better than expected for 4 (25%) and 5 (31%) cardiac hospitals for AMI and CHF, respectively, compared with 5 (4%) and 6 (5%) general hospitals. Risk-standardized mortality rates were modestly lower at cardiac hospitals (15.0% versus 16.2% for AMI, P<0.001, and 10.7% versus 11.3% for CHF, P<0.01).
Conclusions— Patients with AMI and CHF at cardiac hospitals differ considerably from those at peer general hospitals. Although outcomes were modestly better at cardiac hospitals, substantial variation was noted across individual facilities.
Key Words: cardiac care facilities myocardial infarction heart failure, congestive outcomes research
| Introduction |
|---|
|
|
|---|
Editorial p 2238
Clinical Perspective p 2287
Selectively focusing on specific medical procedures or diseases may allow specialty hospitals to provide better quality of care by concentrating physician expertise and hospital resources, but earlier studies examining this issue have reported mixed results.5,6 More concerning, these reports primarily have investigated the performance of these hospitals with regard to low-risk patients undergoing procedures. Less is known about the ability of specialty hospitals to care for broader populations of hospitalized patients, particularly those with medical conditions that require complex management and who may not necessarily require procedures. With an anticipated growth in the number of specialty hospitals and their potential for an expanded role in the US healthcare system, it becomes critical to understand how these facilities perform in these populations, especially for key quality measures such as risk-adjusted mortality rates.
Accordingly, we compared the performance of specialty cardiac hospitals with "peer" general hospitals among Medicare patients with 2 common conditions that are associated with substantial morbidity and mortality in the United States: acute myocardial infarction (AMI) and congestive heart failure (CHF). We focused on cardiac hospitals because they receive the majority of Medicare payments to specialty hospitals and typically are less procedurally focused than specialty surgical or orthopedic hospitals.7 We were also interested in better characterizing patients with these conditions treated at cardiac hospitals. To evaluate hospital performance, we used validated 30-day mortality risk models for AMI and CHF recently endorsed by the National Quality Forum and in use for public reporting by CMS.8
| Methods |
|---|
|
|
|---|
|
|
We identified cardiac hospitals using the recent "Report to Congress on the Strategic and Implementing Plan for Specialty Hospitals" released by CMS.4 This report initially identified specialty hospitals using Medicare Payment Advisory Committee criteria and was supplemented by a list of facilities that had recently requested an advisory opinion by CMS as to their status as a specialty hospital.11 From the CMS report, we identified a total of 16 cardiac hospitals providing data on care to Medicare beneficiaries in 2003. The present comparison group of 121 "peer" general hospitals was limited to those facilities that performed coronary revascularization (coronary artery bypass grafting [CABG] or percutaneous coronary intervention) within the same healthcare market. We required that all peer general hospitals perform coronary revascularization given that these clinical services were available in all cardiac hospitals. Healthcare markets were defined by use of hospital referral regions from the Dartmouth Atlas of Health Care and were based on the facilitys ZIP code.12
To measure patient variables, additional data were obtained from diagnosis codes obtained 12 months before the index admission from Medicare part A inpatient and outpatient data. These data include facility claims for emergency department visits, ambulatory surgery, and surgical or diagnostic procedures. Medicare part B data, which include provider encounters, were also used for patient information. For patients transferred from 1 facility to another, both hospitalizations were linked into a single episode of care, with outcomes attributed to the first hospital (regardless of whether the hospital was a cardiac hospital or peer general hospital) and comorbidities identified only from the first hospitalization to avoid the misclassification of complications as preexisting conditions.
Patients discharged alive, not against medical advice, and within 24 hours were excluded because such patients were unlikely to have had an AMI or decompensated CHF. We also excluded patients admitted to hospitals with 11 or fewer cases for that diagnosis during the study period. Because the regression model incorporates utilization information 12 months before the index hospitalization, only patients enrolled in Medicare fee-for-service for at least 1 year were included.
Outcomes, Risk-Adjustment Variables, and Statistical Analyses
The primary outcome was death due to any cause within 30 days after the index admission date, which was obtained from Medicare enrollment files. For AMI and CHF, the patient variables used for risk adjustment included the demographic, cardiovascular, and comorbidity variables previously described in the tables above. We compared baseline characteristics and outcomes of patients by hospital type using
2 tests for categorical variables and ANOVA for continuous variables.
With data from the 137 cardiac and peer general hospitals included in the present study, we used hierarchical generalized linear models to calculate hospital-specific, standardized mortality ratios (SMRs) and risk-standardized mortality rates, accounting for patient characteristics and differences within and between hospitals (to address potential effects of clustering).13 SMRs >1 indicate that a hospital had higher than expected mortality, and <1 indicates lower than expected mortality. To generate the 95% confidence interval estimates of SMRs and the variance of risk-standardized mortality rates for each hospital, a bootstrapping algorithm of N from N patients within each institution with 1000 replications was used.
To compare individual hospital performance between cardiac and peer general hospitals, each hospital was categorized on the basis of the 95% interval estimates of the SMRs into 1 of 3 groups: predicted mortality significantly better than expected mortality, predicted mortality no different than expected mortality, or predicted mortality significantly worse than expected mortality. Distributions of the numbers of cardiac and peer general hospitals within these 3 groups were then compared. Given that some cardiac hospitals do not provide emergency services, which leads to bias in the types of patients admitted, we conducted a sensitivity analysis that excluded hospitals without these services.
Risk-standardized mortality rates, which essentially transform the mortality ratios to rates, were also estimated for cardiac and peer general hospitals stratified by hospital referral region. To account for the fact that some of the risk-standardized mortality rates were estimated more precisely than others, we weighted each rate by its precision (ie, by the inverse of its variance). Differences within referral regions were computed and then combined across hospital referral regions into an overall difference between cardiac hospitals and peer hospitals by use of the weighted estimator.
All analyses were undertaken for AMI and CHF separately and were conducted with SAS software, version 9.1 (SAS Institute, Inc, Cary, NC) with the hierarchical models generated with the GLIMMIX procedure.
The authors had full access to and take full responsibility for the integrity of the data. All authors have read and agree to the manuscript as written.
| Results |
|---|
|
|
|---|
The mean number of patients admitted to cardiac hospitals was similar to peer general hospitals for AMI (119.5 [SD 71.9] versus 108.7 [SD 73.6]; P=0.58) but significantly lower for CHF (79.7 [SD 50.4] versus 151.2 [SD 86.2]; P=0.002). In contrast, cardiac hospitals had higher annual volumes of percutaneous coronary interventions (497.1 [SD 251.1] versus 194.2 [SD 168.7]; P<0.001) and coronary artery bypass grafting procedures (238.3 [SD 154.5] versus 81.7 [SD 84.8]; P<0.001).
We found important differences in baseline characteristics between patients admitted to cardiac hospitals and general hospitals. Compared with those admitted to peer general hospitals, patients with AMI at cardiac hospitals were younger, were more likely to be male, and had a lower prevalence of cardiovascular diseases (Table 2). Patients with CHF (n=1275) admitted to cardiac hospitals were also younger, were more likely to be male, and had a higher prevalence of cardiovascular diseases (Table 3). Most noticeably, patients with AMI and CHF at peer general hospitals had much higher rates of noncardiovascular diseases in general, with nearly 2- to 3-fold higher rates of malnutrition, functional disability, dementia, and major psychiatric disorders.
|
Mean unadjusted 30-day mortality rates were significantly lower in cardiac hospitals than in peer general hospitals for AMI (10.0% versus 16.9%, P<0.001) and CHF (7.1% versus 11.9%, P<0.001). Although these differences were attenuated substantially after adjustment for patient characteristics, risk-standardized mortality rates remained lower in cardiac hospitals for both conditions (15.0% versus 16.2% for AMI, P<0.001, and 10.7% versus 11.3% for CHF, P<0.001; Table 4). The exclusion of cardiac hospitals without emergency care services did not change these findings, with risk-standardized mortality rates still lower in cardiac hospitals for both conditions (14.8% versus 16.2% for AMI, P=0.002, and 10.7% versus 11.4% for CHF, P=0.011).
|
Of note, risk-standardized mortality rates varied widely across both groups of hospitals. For AMI, risk-standardized mortality rates varied from 12.8% to 16.7% for cardiac hospitals and from 13.4% to 19.8% for peer general hospitals. For CHF, risk-standardized mortality rates varied from 8.8% to 12.7% for cardiac hospitals and from 8.5% to 15.3% for peer general hospitals. Distributions are represented in a side-by-side comparison in Figures 1 and 2
, with extremes at peer general hospitals, expressed by quartiles, shown to highlight patterns of performance at cardiac hospitals. No significant differences in the performance of peer general hospitals between quartiles should be assumed.
|
|
Four (25%) and 5 (31%) cardiac hospitals had predicted mortality rates significantly lower than their expected mortality rates (ie, SMRs <1.0) for AMI and CHF, respectively, and none had predicted mortality rates significantly higher than their expected mortality rates (Table 4). In comparison, 5 (4%) and 6 (5%) peer general hospitals had predicted mortality rates significantly lower than their expected mortality rates for AMI and CHF, respectively, whereas 7 (6%) peer general hospitals had predicted mortality rates significantly higher than their expected mortality rate for both AMI and CHF.
Stratified differences in risk-standardized mortality rates between cardiac and peer general hospitals within hospital referral regions are displayed in Figures 3 and 4
. These differences indicated that cardiac hospitals performed better in 2 specific hospital referral regions, with lower risk-standardized mortality rates for both AMI (mean difference –1.3%; 95% interval estimates –1.9% to –0.6%) and CHF (mean difference –1.1%; 95% interval estimates –1.7% to –0.5%; Figures 3 and 4
).
|
|
| Discussion |
|---|
|
|
|---|
The finding that patients with medical conditions at cardiac hospitals have lower risk profiles on average than those at peer general hospitals is important and highly consistent with prior studies that have examined quality of care in patients undergoing cardiovascular procedures at these facilities.14 Supporters of cardiac hospitals have argued that selective referral of low-risk patients to these facilities reflects responsible care, because high-risk patients often require multidisciplinary resources that may not be routinely available there. However, critics have pointed out that because cardiac hospitals are "cherry-picking" healthier patients, many general hospitals are left to care for older and more severely ill patients.15 This potentially places competing peer general hospitals at greater financial risk. These issues encourage refinement of the inpatient payment schedule by healthcare purchasers with a particular focus on addressing differences in disease acuity.
The finding of better outcomes on average at cardiac hospitals for AMI and CHF is new information. Previous reports have primarily focused on outcomes at specialty hospitals for medical procedures such as coronary artery bypass grafting and percutaneous coronary intervention.6 Such improvements have largely been attributed to greater provider control over hospital operations and inpatient care, as well as the ability of these facilities to concentrate clinical expertise.11 The present findings are the first (that we are aware of) to indicate potentially better outcomes at cardiac hospitals for medical conditions such as AMI and CHF. As with cardiovascular procedures, a possible reason for these findings may be related to a greater ability of cardiac hospitals to focus expertise among both physicians and ancillary staff. Another explanation may be that physicians who opt to set up cardiac hospitals are systematically better physicians. Finally, it may be that cardiac hospitals (which are typically newer) provide more comprehensive or advanced services, which leads to better outcomes. However, it is noteworthy that better outcomes were not found in many cardiac hospitals, which suggests that the designation does not guarantee a particular level of performance.
Recently, CMS released its final report to Congress implementing a strategic plan for specialty hospitals as required by the Deficit Reduction Act of 2005.4 In this plan, no specific restrictions were placed on the development of new specialty hospitals, thus effectively ending the temporary moratorium on new construction. Instead, CMS primarily focused its attention on revising the inpatient payment schedule in order to "level the playing field" between specialty and general hospitals. This was also done in an effort to limit financial incentives to invest in certain types of services simply because of their profitability. The plan also proposes new methods for implementing "gain sharing" and value-based approaches at general hospitals to better align physician and hospital incentives toward improving care. In addition to reducing the potential for cherry-picking, however, these revisions are likely to increase admissions to cardiac hospitals for medical conditions such as AMI and CHF.
The present findings should be interpreted in the context of the following study design issues. First, the risk-standardized mortality models used in the present study relied on administrative data, and all of the clinical factors that possibly contribute to variation in mortality may not be accounted for, which could result in the potential for residual confounding. However, our use of national Medicare inpatient and outpatient files allowed us to exhaustively capture clinical and utilization information in these patients over the study period. In addition, these models have been shown to produce estimates that are similar to those from models based on chart-based data and have been endorsed by expert panels.8 Second, cardiac hospitals may be associated with better nonmortality outcomes, such as improvements in patient satisfaction. Prior studies support the potential benefits of specialty hospitals in this regard, primarily because of the improved amenities often available at these facilities.11 Third, we identified only 16 cardiac hospitals, and many admitted few patients with AMI and CHF (as did some peer general hospitals). This could have limited our statistical power to detect important differences in hospital performance in some regions. However, we did exclude hospitals with very small numbers of patients, "weighted" the comparisons of hospital performance, and used hierarchical models for calculating risk-standardized mortality rates. The later allowed us to "borrow" power across hospitals when we directly compared facilities.
We found important differences between patients admitted with AMI and CHF at cardiac hospitals compared with those at peer general hospitals within the same healthcare market. After adjustment for these differences in patient characteristics, cardiac hospitals had lower risk-standardized mortality rates for both conditions than peer general hospitals, although considerable variation was present across individual facilities. As the number of cardiac hospitals continues to grow in the United States, their role in caring for patients with these conditions is likely to increase, particularly if changes in the inpatient reimbursement schedule begin to shift away from medical procedures. Determining the particular practice patterns at cardiac hospitals that are associated with better outcomes in patients with AMI and CHF and transferring these approaches to peer general hospitals may become an important strategy for improving overall quality of care for these conditions.
| Acknowledgments |
|---|
The analyses on which this publication is based were performed under contract No. HHSM-500-2005-CO001C, entitled "Utilization and Quality Control Quality Improvement Organization for the State (Commonwealth) of Colorado," sponsored by the Centers for Medicare & Medicaid Services, an agency of the US Department of Health and Human Services. This project was supported in part by a grant from the Agency for Healthcare Research and Quality (1R01HS015571-01A1). Dr Ross is supported by the Department of Veterans Affairs Health Services Research and Development Service, project No. TRP-02-149. The agencies and foundations that funded this work were not involved in the design and conduct of the study, in data management or analysis, or in manuscript preparation.
Disclosures
None.
| References |
|---|
|
|
|---|
| Footnotes |
|---|
The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, the Department of Veterans Affairs, or the Agency for Healthcare Research and Quality, 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.
Guest Editor for this article was James E. Udelson, MD.
This article has been cited by other articles:
![]() |
H. H. Ting, H. M. Krumholz, E. H. Bradley, D. C. Cone, J. P. Curtis, B. J. Drew, J. M. Field, W. J. French, W. B. Gibler, D. C. Goff, et al. Implementation and Integration of Prehospital ECGs Into Systems of Care for Acute Coronary Syndrome: A Scientific Statement From the American Heart Association Interdisciplinary Council on Quality of Care and Outcomes Research, Emergency Cardiovascular Care Committee, Council on Cardiovascular Nursing, and Council on Clinical Cardiology Circulation, September 2, 2008; 118(10): 1066 - 1079. [Full Text] [PDF] |
||||
![]() |
E. R. Bates and B. K. Nallamothu Commentary: The Role of Percutaneous Coronary Intervention in ST-Segment-Elevation Myocardial Infarction Circulation, July 29, 2008; 118(5): 567 - 573. [Full Text] [PDF] |
||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Circulation Home | Subscriptions | Archives | Feedback | Authors | Help | AHA Journals Home | Search Copyright © 2007 American Heart Association, Inc. All rights reserved. Unauthorized use prohibited. |