|
Quality Of Care For Acute Myocardial Infarction At Urban Safety-Net Hospitals
Safety-net hospitals are experiencing increasing financial strains, possibly affecting their quality of care. We compare quality at safety-net and non-safety-net urban hospitals for Medicare beneficiaries admitted with acute myocardial infarction (AMI). Although safety-net hospitals had modestly higher risk-standardized thirty-day all-cause mortality rates and modestly lower adherence to quality-of-care performance measures than non-safety-net hospitals, there was much heterogeneity among safety-net hospitals and substantial overlap with non-safety-net hospitals. We examine the implications of these findings for the millions of vulnerable Americans who rely on safety-net hospitals for their care.
THE U.S. HEALTH CARE SYSTEM has long maintained a patchwork system of physicians and hospitals, known as the safety net, that is committed to caring for populations without stable access to care: the uninsured, the low-income underinsured, and Medicaid beneficiaries.1 Recent policy and market changes are likely to have increased financial strain among all hospitals; however, among safety-net hospitals, this strain has been related to the rising number of un-insured adults requiring indigent care, the growth of health maintenance organizations (HMOs) in the Medicaid sector, and the Balanced Budget Act (BBA) of 1997.2 Specifically, the BBA decreased Medicare payments by limiting inpatient diagnosis-related group (DRG) inflation adjustments, reducing medical education payments and disproportionate-share hospital (DSH) payment adjustments, and changing payment methods for capital expenses and outpatient services, in addition to establishing Medicaid DSH spending limits.3 Partially as a result of these financial stresses, safety-net hospitals have closed in some areas.4 Many of those remaining open are further burdened as non-safety-net hospitals trim services used by the indigent, including emergency department (ED) services, AIDS care, and in- and outpatient substance abuse care.5 The Institute of Medicine (IOM) has declared the nation’s safety net to be "intact but endangered."6 Given the financial strain on safety-net hospitals and the associations between lower quality of care and both worse baseline financial health and increased fiscal strain, particularly for vulnerable populations, we sought to evaluate the quality of care delivered at safety-net and non-safety-net hospitals for patients admitted for acute myocardial infarction (AMI), using (1) risk-standardized thirty-day all-cause mortality and (2) quality-of-care performance measures.7 We restricted our analysis to urban areas where there were both safety-net and non-safety-net hospitals, to minimize any composition bias that would be introduced by including non-safety-net hospitals from urban areas without safety-net hospitals. Also, because differences in coverage and access have been associated with differences in quality of care, we focused only on Medicare beneficiaries, an insured population with generally stable access to care, thereby minimizing the confounding of differences between safety-net and non-safety-net hospitals from unobservable differences between populations receiving care at each set of hospitals.8
Data. For this study we used the 2003 Medicare Provider Analysis and Review (MEDPAR) data, the Centers for Medicare and Medicaid Services (CMS) Hospital Quality Alliance (HQA) data for quarters 3 and 4, and 2002 American Hospital Association (AHA) Annual Survey data. We included all fee-for-service (FFS) Medicare beneficiaries age sixty-five and older who were hospitalized with AMI. Patients who were transferred between acute care facilities were assigned to the referring hospital. We excluded patients whose insurance coverage during 2003 crossed over between FFS Medicare and any HMO, with fewer than twelve months of continuous Medicare FFS enrollment, for whom AHA data were not available, or who were hospitalized outside of the United States or in nonurban areas, as defined by U.S. Census Bureau metropolitan statistical areas (MSAs). To further reduce heterogeneity, we excluded hospitals in MSAs that did not contain at least one safety-net and one non-safety-net hospital, hospitals that reported fewer than five AMI hospitalizations throughout 2003, or hospitals whose safety-net status could not be determined (one hospital). A total of 116,857 patient hospitalizations in 2,218 hospitals met one or more of these criteria; the remaining 151,712 hospitalizations in 1,645 hospitals within 169 MSAs constitute the study cohort. However, because we were unable to obtain the CMS Q3/Q4 HQA data for another 1,692 patients in 68 hospitals, our study cohort for the examination of quality-of-care performance measures was a slightly smaller 150,020 hospitalizations in 1,577 hospitals within 169 MSAs. Variables. Safety-net hospitals were defined using an inclusive definition that has been used in previous research: either public hospitals or private hospitals with a Medicaid caseload greater than one standard deviation above their respective state’s mean private hospital Medicaid caseload using 2002 AHA Annual Survey data.9 Hospital quality was defined using both outcome and process measures. First, we determined hospital-specific risk-standardized thirty-day all-cause mortality rates (RSMRs) after admission for AMI using a risk-standardization model that includes information available in administrative data.10 The model produces estimates of RSMRs after admission for AMI that are good surrogates for estimates from a medical-record model; it has been endorsed by the National Quality Forum. It includes patient-specific information on age and sex, ten features from the past cardiovascular history, and fifteen other comorbid conditions by using hierarchical condition categories.11 Second, using 2004 CMS Q3/Q4 HQA data, we created two composite rates using eight hospital-specific quality-of-care performance measures for AMI hospitalizations that are widely endorsed as valid and reflective of high-quality hospital care.12 The eight performance measures were categorized as either nonreperfusion or time-to-reperfusion measures. The former included aspirin at arrival, aspirin prescribed at discharge, beta-blocker at arrival, beta-blocker prescribed at discharge, angiotensin-converting enzyme (ACE) inhibitor or angiotensin-receptor blocker (ARB) for left ventricular systolic dysfunction, and smoking cessation advice or counseling. The latter included fibrinolytic agent received within thirty minutes of hospital arrival and percutaneous coronary intervention received within 120 minutes of hospital arrival. A composite rate was then created for the two sets of performance measures, each based on the percentage of opportunities fulfilled. The composite rate methodology was developed as part of the Premier and CMS Hospital Quality Incentive Demonstration (HQID) and is a valid and reliable method to aggregate data and compare hospitals using a single measure.13 Analysis. Using chi-square and t-test analyses, we compared safety-net and non-safety-net hospitals for patient-level differences in sociodemographic characteristics, cardiovascular history, and comorbid conditions, as well as for hospital-level differences in hospital and MSA characteristics. We then developed two-level (hospital and MSA) hierarchical linear regression models to examine the association between safety-net status and median and mean hospital-specific RSMRs and performance measure adherence rates (PMARs). In the model analyses, the first-level model specification included hospital characteristics (except ownership status and Medicaid caseload), and the second-level specification included MSA size (categorized). In all hierarchical models we considered MSA-level random effects to account for the clustering (nonindependence) of hospitals within an MSA. All models were repeated among only private hospitals (for-profit and nonprofit) to examine the association between private safety-net hospitals and quality of care. To facilitate the interpretation of our results, we calculated the hospital-specific RSMRs and PMARs from corresponding models. Means and medians of the unadjusted and adjusted hospital-specific rates were compared between safety-net and non-safety-net hospitals using t-tests and nonparametric median score tests. We repeated our analyses to include patients’ race. The results from this sequence of analyses were no different, so we present only the results from the analyses described in detail above, because they were derived using our validated model. Statistical analyses were conducted using SAS version 9.1 and HLM 5.04. All statistical tests were two-tailed.14
More than 32,000 of the patient hospitalizations we examined were in 455 safety-net hospitals, accounting for 21 percent of hospitalizations and 28 percent of hospitals. The mean age of all patients included in our cohort was seventy-nine years, and 52 percent were female (Exhibit 1  ). Patients admitted to safety-net hospitals were younger than those admitted to non-safety-net hospitals and more likely to have a history of comorbid conditions, including prior AMI, heart failure, unstable angina, diabetes, and renal failure. Among safety-net hospitals, 223 were public, 178 were not-for-profit, and 54 were for-profit (Exhibit 2). In addition, safety-net hospitals were more likely to be teaching hospitals and had lower mean AMI volume but were not significantly less likely to provide either coronary artery bypass graft (CABG) surgery or cardiac catheterization.
Safety-net hospitals and RSMRs. In unadjusted analyses, safety-net hospitals were associated with significantly higher median hospital-specific RSMRs compared with non-safety-net hospitals (Exhibit 3 ). After adjustment, the median hospital-specific RSMRs remained significantly different (Exhibit 4). Relative differences in mean hospital-specific RSMRs between safety-net and non-safety-net hospitals within MSAs varied widely (Exhibit 5), such that 19.5 percent of safety-net hospitals had an RSMR greater than 2 percent lower; 40.8 percent, between 2 percent lower or higher; and 39.6 percent, greater than 2 percent higher than those of non-safety-net hospitals within their respective MSA.
Safety-net hospitals and PMARs. In unadjusted analyses, there was no difference in median hospital-specific nonreperfusion PMARs between safety-net and non-safety-net hospitals (Exhibit 3  ), although safety-net hospitals were associated with significantly lower median hospital-specific time-to-reperfusion PMARs compared with non-safety-net hospitals (Exhibit 3). After adjustment, safety-net hospitals were associated with slightly but, significantly, lower median hospital-specific PMARs for both sets of measures compared with non-safety-net hospitals (Exhibit 4). Relative differences in mean hospital-specific PMARs between safety-net and non-safety-net hospitals within MSAs varied widely for both sets of measures (Exhibit 5), such that for nonreperfusion performance measures, 25.4 percent of safety-net hospitals had a PMAR greater than 2 percent higher; 43.2 percent, between 2 percent higher or lower; and 31.4 percent, greater than 2 percent lower than those of non-safety-net hospitals within their respective MSAs. For time-to-reperfusion process measures, 20.1 percent had a PMAR greater than 2 percent higher; 22.5 percent, between 2 percent higher or lower; and 57.4 percent, greater than 2 percent lower. Private safety-net hospitals and quality of care. In adjusted analyses, private safety-net hospitals were associated with significantly higher median RSMRs (22.0 percent versus 20.2 percent; p < .001) and slightly, but significantly, lower PMARs (nonreperfusion: 88.2 percent versus 89.2 percent [p < .001]; time to reperfusion: 45.5 percent versus 49.3 percent [p < .001]) when compared with private non-safety-net hospitals (data not shown).
We found that urban safety-net hospitals overall had modestly higher RSMRs and modestly worse quality-of-care PMARs for Medicare beneficiaries admitted for AMI compared with urban non-safety-net hospitals. With regard to the performance measures, although the absolute difference was larger for time-to-reperfusion than for nonreperfusion measures, differences are small compared with the difference between either set of measures and ideal care. However, the 1.9 percentage points greater median hospital-specific RSMR after admission for AMI translates to one excess death for approximately every fifty-four AMI admissions at safety-net compared with non-safety-net hospitals. Heterogeneity in performance. There was much overlap in quality of care between safety-net and non-safety-net hospitals, which indicates that despite the challenges these safety-net hospitals face, many perform as well as or better than non-safety-net hospitals. Nearly a fifth of safety-net hospitals had lower RSMRs than the non-safety-net hospitals within their MSAs. Moreover, approximately the same number of safety-net hospitals had higher PMARs than the non-safety-net hospitals within their MSAs. Thus, our findings must be tempered by the marked heterogeneity in quality of care for AMI among safety-net hospitals. Our finding of heterogeneity in performance indicates the need for continued efforts to assess the quality of hospital care, perhaps as part of the National Healthcare Disparities Report.15 Moreover, initiatives such as Safety Net Monitoring, sponsored by the Agency for Healthcare Research and Quality (AHRQ), need to be supported so that the quality of safety-net hospitals’ care can be easily monitored.16 Interestingly, we found that at the farthest extremes, the safety-net hospitals that most outperformed or underperformed the non-safety-net hospitals within their MSAs with regard to RSMRs tended to do so for PMARs as well. For instance, safety-net hospitals within MSAs in Massachusetts, Mississippi, Florida, Connecticut, and New York were the most successful comparatively, while safety-net hospitals within MSAs in Minnesota, Florida, and Ohio were the least successful comparatively. Using monitoring initiatives, high-quality hospitals can be identified and subsequently examined in an effort to determine which institutionalized systems and processes of care enable and sustain the provision of high-quality care. In addition, low-quality hospitals can be identified and targeted to ensure support and improvement in hospitals and MSAs with the most pressing needs. Depending on local requirements and demands, state or local investments, federal financial assistance, or other monetary incentives could be used to maintain or improve quality of care at safety-net hospitals. Importance of continued funding. Furthermore, programs that support high-quality care for safety-net hospitals need to be funded and endorsed, and the implications of reducing support for safety-net hospitals for vulnerable populations should be appreciated. For instance, the Healthy Community Access Program (HCAP), a federal program directed by the U.S. Department of Health and Human Services (HHS), has invested more than $400 million since 2000 in strengthening safety-net hospitals by providing resources to develop information systems or create language access programs.17 However, the Departments of Labor, Health and Human Services, and Education and Related Agencies Appropriations Act of 2006, passed in December 2005, eliminated funding for HCAP for fiscal year 2006, and funding for FY 2007 is unlikely. In addition, the presidential FY 2007 budget proposes cutting $36 billion in Medicare hospital payments over the next five years and $35 billion in federal Medicaid support over the next ten years. Safety-net hospitals are already known to be experiencing substantial fiscal strain, and these proposed budget cuts will further strain, if not break, our patchwork system. These budget changes, if enacted, need to be accompanied by robust efforts to assess their impact on the quality of care at safety-net hospitals to both ensure their continued presence and operation and improve the quality of their care. Study limitations. By examining Medicare beneficiaries only, we restricted our analysis to adults regularly using safety-net hospitals who have the best coverage and access. The IOM has focused on community effects and commented on shared destiny: how the quality, quantity, and scope of health services within the community can be adversely affected by having a large or growing uninsured population.18 The lower quality that we observed at some safety-net hospitals is likely to affect all hospitalized adults, regardless of insurance coverage or income, akin to the center effect that has been described in research examining racial disparities.19 Our study is one of the first to examine the quality of care at safety-net hospitals. However, several considerations must be kept in mind as one interprets its results. We examined only one condition, AMI, so our results might not be generalizable to quality for other conditions. We also used a very broad definition of safety-net hospitals, which could be overly inclusive. However, in so doing, we have conservatively estimated the differences in quality of care delivered at safety-net compared with non-safety-net hospitals. In addition, the association between safety-net status and lower quality remains strong and significant even when the analysis includes only private hospitals, which demonstrates that the inclusion of all public hospitals in our definition is not confounding the results. Finally, we cannot rule out that the differences we found in quality-of-care PMARs are the result of more-comprehensive coding at non-safety-net hospitals. OUR STUDY DEMONSTRATES THAT MODEST, but clear, differences exist between the quality of AMI care delivered at urban safety-net and non-safety-net hospitals, on average, despite considerable heterogeneity among safety-net hospitals and substantial overlap with non-safety-net hospitals across the MSAs examined. Efficient and effective support of the safety net, along with ongoing surveillance, is imperative for ensuring its continued presence and operation, as well as for increasing the overall provision of high-quality care.
Joseph Ross is an instructor in the Department of Geriatrics and Adult Development, Mount Sinai School of Medicine, in New York City. Stephen Cha is an assistant professor (adjunct) in the Department of Medicine, Yale University School of Medicine, in New Haven, Connecticut. Andrew Epstein is an assistant professor of public health in the Department of Epidemiology and Public Health; Yongfei Wang is a lecturer in the Department of Medicine; Elizabeth Bradley is a professor of public health in the Department of Epidemiology and Public Health; Jeph Herrin is an assistant professor of medicine (adjunct) in the Department of Medicine; and Judith Lichtman is an assistant professor of public health in the Department of Epidemiology and Public Health. Sharon-Lise Normand is a professor of health care policy (biostatistics) in the Department of Health Care Policy, Harvard Medical School, in Boston, Massachusetts. Frederick Masoudi is a cardiologist in the Department of Medicine (Cardiology), Denver Health Medical Center, in Denver, Colorado. Harlan Krumholz (harlan.krumholz{at}yale.edu) is the Harold H. Hines Jr. Professor of Medicine and Epidemiology and Public Health at Yale University School of Medicine. A previous version of this paper was presented at the 2006 Society of General Internal Medicine Annual Meeting, Los Angeles, California, 27 April 2006. Harlan Krumholz has research contracts with the American College of Cardiology and the Colorado Foundation for Medical Care; is on the advisory boards of Amgen, Alere, and UnitedHealthcare; serves as a subject matter expert for VHA Inc.; and is editor-in-chief of Journal Watch Cardiology of the Massachusetts Medical Society. Joseph Ross and Stephen Cha were scholars in the Robert Wood Johnson Clinical Scholars Program at Yale University, sponsored by the Robert Wood Johnson Foundation (RWJF) and the U.S. Department of Veterans Affairs (VA), respectively, at the time of their project involvement. Neither the RWJF nor the VA had any role in the design or conduct of the study; collection, management, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript. Elizabeth Bradley is supported by a Patrick and Catherine Weldon Donaghue Medical Research Foundation Investigator Award. The Donaghue Foundation had no role in the design or conduct of the study; collection, management, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript. Judith Lichtman is supported by Grant no. 1 K01 DP000085-03 from the U.S. Centers for Disease Control and Prevention (CDC). The contents of this manuscript are solely the responsibility of the authors and do not necessarily represent the official views of the CDC.
This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Top
Study Data And Methods
Study Results
