Effect of Cuts in Medicare Reimbursement on Process and Outcome of Care for Acute Myocardial Infarction Patients
http://www.100md.com
《循环学杂志》
Philadelphia Veterans Affairs Medical Center (K.G.V.), and School of Medicine (K.G.V., J.Z.) and the Wharton School (K.G.V.), University of Pennsylvania, Philadelphia
University of Chicago (R.T.K.), Chicago, Ill
Ovation Research Group (L.P.), Seattle, Wash
Duke University (E.P.), Durham, NC.
Abstract
Background— The Balanced Budget Act (BBA) of 1997 was designed to reduce Medicare reimbursements by $116.4 billion from 1998 to 2002. The objective of this study was to determine whether the process of care for acute myocardial infarction (AMI) worsened to a greater degree in hospitals under increased financial strain from the BBA and whether vulnerable populations such as the uninsured were disproportionately affected.
Methods and Results— We examined how process-of-care measures and in-hospital mortality for AMI patients changed in accordance with the degree of BBA-induced financial stress using data on 236 506 patients from the National Registry of Myocardial Infarction (NRMI) and Medicare Cost Reports from 1996 to 2001. BBA-induced reductions in hospital net revenues were estimated at 1.5% ($2.9 million) for hospitals with low BBA impact and 3.2% ($3.7 million) for hospitals with a high impact in 1998, worsening to 2.2% ($4.4 million) and 4.7% ($6.0 million), respectively, by 2001. For both insured and uninsured patients in high- versus low-impact hospitals, there was no systematic worsening of time to thrombolytic therapy, balloon inflation, medication use on admission, medication use on discharge, or mortality. There was no systematic pattern of different treatment among the insured and uninsured. Operating margins decreased to a degree commensurate with the degree of revenue reduction in high- versus low-impact hospitals.
Conclusions— BBA created a moderate financial strain on hospitals. However, process-of-care measures for both insured and uninsured patients with AMI were not appreciably affected by these revenue reductions. It is important to note that these results apply only to AMI patients; we do not know the degree to which these findings generalize to other conditions.
Key Words: delivery of health care health policy outcome assessment (health care)
Introduction
The Balanced Budget Act (BBA) of 1997 was designed to reduce Medicare reimbursements to hospitals, physicians, home health agencies, and skilled nursing facilities by $116.4 billion from 1998 to 2002.1 As a result of significant protests from teaching hospitals and the American Hospital Association, which cited "service closures and cutbacks,"2,3 Congress passed 2 laws to restore a portion of these cuts: the Balanced Budget Refinement Act (BBRA) in 1999, which was expected to restore $8.4 billion, and the Benefits and Improvement Protection Act (BIPA) in 2000, which was expected to restore a further $11.5 billion of the original $116 billion reductions.
The impact of reimbursement policy changes on quality of care has been seen in state-level analyses. The uninsured, whose care is funded largely from hospital margins, appears to be the most vulnerable group. Price-competitive market reforms in New Jersey in the early 1990s were associated with rises in inpatient mortality rates for uninsured but not insured acute myocardial infarction (AMI) patients and with concomitant decreases in the rates of revascularization and cardiac catheterization.4 Similarly, the enactment of legislation in California to permit selective contracting for insurance companies (thereby allowing them to price-shop among hospitals) decreased the quantity of services provided to the uninsured.5
We focus on AMI patients because cardiac care tends to represent a meaningful proportion of hospital budgets and patients and therefore is likely to be affected by financial pressures. In contrast to many other conditions, AMI patients are always admitted to hospitals, mitigating the importance of selection bias resulting from varying admission thresholds in studies of changes in hospital quality over time. Numerous studies have examined the effects of policies on AMI mortality for this reason, but few have examined process measures of quality, which may be more sensitive to changes in hospital finances than mortality and thus more reflective of factors under the hospital’s control. Detailed process measures of quality are available for AMI care but not for other conditions. It has become increasingly clear that systems factors have a great influence on quality of care,6,7 and such systems can be expensive to design and implement. The rate of provision of high-quality systems to ensure proper provision of medications and timely delivery of cardiac interventions may thus be affected by hospital financial stress.
We used data from the National Registry of Myocardial Infarction (NRMI) from 1996 to 2001 to examine whether quality of care worsened to a greater degree over time in hospitals under more financial strain from the BBA. We examined a set of measures known to be markers of quality of care for AMI patients8: in-hospital mortality; time to treatment with either primary angioplasty or thrombolytic therapy for patients with ST-segment elevation AMI; and rates of use of aspirin, -blockers, and ACE inhibitors within 24 hours of admission and on discharge. We focused on whether outcomes for the uninsured worsened to a greater degree than for the insured because cost-saving reforms such as the BBA could be a mechanism contributing to larger disparities in health outcomes.
Methods
This project was approved by the University of Pennsylvania Institutional Review Board.
Data
The NRMI is a database sponsored by Genentech that collects detailed data on process and outcomes of care of AMI patients from the approximately one third of all hospitals throughout the United States that volunteer to participate. NRMI includes data on hospitals from every state in the country. These data include information on time from symptom onset to arrival in the hospital, receipt of medications within 24 hours of admission and on discharge, time from hospital arrival to ordering and initiation of thrombolytic therapy, and discharge status. We limited our sample to hospitals that contributed at least 20 AMI cases per year to the NRMI registry from 1996 to 2001 because observed changes in process may not be meaningful when very few cases are addressed. We used data from the Medicare Cost Reports for information on net revenues and hospital operating margins and data from the American Hospital Association (AHA) Annual Survey for information on hospital characteristics.
Data Sample
Our initial sample included all interventional hospitals (hospitals with on-site angioplasty [PTCA] capabilities) that participated in the NRMI. The initial data set contained 567 546 admissions for AMI from 1996 to 2001 in 368 hospitals. We excluded hospitals with missing AHA data or unreliable or missing Medicare Cost Report information (88 hospitals; n=145 966); patients with unknown race (n=2149); patients who were transferred to other hospitals, because their ultimate discharge disposition was unknown (n=16 782); and patients transferring from other hospitals, because our focus was on "de novo" cases of AMI (n=125 307). After the above exclusions, hospitals that no longer had at least 20 admissions per year were excluded (72 hospitals; n=40 836). The final data set had 236 506 patients from 208 hospitals; 95 545 patients from 208 hospitals were admitted with ST-segment elevation AMI or left bundle-branch block, and time to thrombolytic therapy and time to balloon inflation were measured only for these patients. The sample of excluded patients did not differ systematically from the included sample. Because differential rates of change in transfer rates in high- and low-impact hospitals could bias the sample, we examined whether transfer rates changed at similar rates for high- and low-impact hospitals from before BBA (1996–1997) to after BBA (1998–2001). Among all hospitals that admitted at least 20 patients per year, transfer rates increased from 28.8% before BBA to 30.0% after BBA among hospitals in the high-impact quartile compared with an increase from 29.0% to 30.2% in the low-impact quartile. The degree of change was clearly quite similar between these 2 groups.
Calculation of the BBA Impact
We used a simulator constructed by the AHA as a tool for hospitals to calculate the financial impact of BBA and BBRA.9 We did not include BIPA in this simulation because these changes were instituted starting in 2001 and had a small financial impact in that year. Using financial information from the 1997 Medicare Cost Reports, the worksheet applies trend factors to simulate hospital Medicare reimbursements without the BBA (ie, under inflation), under the BBA, and under the BBRA. Subtracting the without-BBA estimations from the BBA and BBRA estimations provides a prediction of the financial impacts of the Medicare changes on reimbursements from fiscal 1998 to fiscal 2001 for each hospital. The worksheet includes most aspects of the BBA impact on hospitals, including diagnosis-related group payments, indirect medical education payments, and disproportionate share payments. Unlike measures such as operating margin and net revenue, it avoids contamination of the measure of financial impact by behavioral responses such as diagnosis-related group creep, whereby hospitals may respond to a change in financial reimbursement by altering reporting and charging,10 or adjustments in hospital payer mix. To obtain an estimate of the effect of the BBA on total hospital net revenues, we multiplied the percent change in Medicare reimbursements by the percentage of net patient revenue from Medicare reimbursements in the baseline year of 1997. We used an average BBA impact for each hospital over the 4 years of 1998 through 2001 to divide hospitals into high-impact hospitals (with the highest quartile of impact), low-impact hospitals (with the lowest quartile of BBA impact), and middle-impact hospitals (Table 1).
Adjustment for Patient and Hospital Characteristics
The NRMI includes extensive data on patient characteristics at admission that were used for risk adjustment: patient age, gender, and race; personal history of smoking, hypertension, hypercholesterolemia, congestive heart failure, stroke, angina, angioplasty, or CABG; family history of coronary artery disease; Killip congestive heart failure class; and time from symptom onset to arrival in the hospital. We also adjusted for blood pressure on admission using dummy variables for systolic blood pressure <60, 60 to 80, 81 to 100, and >100 mm Hg and for pulse on admission <50, 51 to 100, and >100 bpm.
On the basis of literature examining the link between hospital characteristics and quality of care,11–16 we adjusted for hospital characteristics that may influence health outcomes: bed size, teaching status (member of the Council of Teaching Hospitals), urban or rural setting, and availability of technology (onsite open heart surgery capabilities). We also examined whether nurse staffing ratios (ratio of RN and LPN to staffed bed) or operating margins changed to a greater degree in high- compared with low-impact hospitals and whether inclusion of these characteristics in our models affected measured changes in process or outcomes measures.
Statistical Analyses
We first examined the degree to which the BBA affected net revenues at different hospitals and divided hospitals into the top 25%, middle 50%, and bottom 25%. We then examined time trends in each study outcome among hospitals in the highest- and lowest-impact groups.
A patient-level logistic regression analysis was used to assess the effect of the BBA on the probability of in-hospital mortality; on receipt of aspirin, -blockers, or ACE inhibitors within 24 hours of admission or on discharge; and on receiving catheterization, angioplasty, or CABG. Linear regression was used to examine the effects of BBA on time to treatment for ST-segment elevation AMI patients. We adjusted for patient and hospital characteristics, intertemporal trends that were common to all hospitals in the sample, and baseline differences in each measure by BBA impact group. The effect of financial stress was identified by use of an interaction term between year and BBA impact group to compare how each study measure changed from before BBA to each of the post-BBA years in the high- compared with low-impact group.
We examined changes in each of our outcome measures separately for the insured and uninsured to test whether, within each of these groups, there were differential changes in the outcome measures for high- and low-impact hospitals. We then combined the data from the insured and uninsured to test whether effect sizes were different among the insured and uninsured. Because the ability of a hospital to respond to the financial stress resulting from Medicare cuts may depend on the amount of slack in financial resources at baseline, we examined the sensitivity of our results to inclusion of baseline operating margins in the regressions. We focused on baseline operating margin as a proxy for the amount of slack in financial resources because extensive economics literature indicates that hospitals trade off between operating margins and quality above the profit-maximizing level in the face of net revenue reductions.5,17 Given this, operating margins at baseline are probably the best proxy for the degree to which these tradeoffs have been made by hospitals before BBA. We further test whether adjusting for fund balances (assets minus liabilities) at baseline affected the quality response to BBA. Finally, we tested whether the BBA affected length of stay among the insured or uninsured or operating margins, examining each of these in turn as dependent variables.
Observations among patients from the same hospital may be correlated, and using aggregate variables such as hospital-wide financial impact in a patient-level regression can potentially lead to overestimation of the significance of such variables. To adjust for these problems, we used regression with robust standard errors clustered on the hospital level.18 For all of our models, we used generalized estimating equations that account for clustering of patients within hospitals. The statistical program used was SAS version 8.02. For both the linear and logistic regressions, the GENMOD procedure was used.
Results
Descriptive Analysis
Medication use within 24 hours of hospital admission and on discharge increased in both the high- and low-impact groups and among both the insured and uninsured (Table 3). Time to thrombolytic therapy ordering and initiation decreased slightly in the low-impact group among the insured (0.1 hours) while remaining constant or increasing slightly among the insured within the high-impact group and remaining constant or decreasing slightly among the uninsured in both impact groups. Door to balloon inflation times decreased more in high-impact hospitals among the insured (0.5 hours) than in low-impact hospitals (0.3 hours). The year-to-year trends for door to balloon time and door to initiation of thrombolytic therapy are illustrated in the Figure. This figure shows that the time to treatment is similar among the insured and uninsured and that there is no clear worsening in high-impact hospitals relative to low-impact hospitals except to a small degree in initiation of thrombolysis among the insured. Mortality declined slightly in both the low- and high-impact groups from before BBA to after BBA in unadjusted analyses except for an increase among the uninsured in the high-impact group.
Changes in time to treatment in high- vs low-impact hospitals over time.
Changes in Procedure Rates in High- Versus Low-Impact Hospitals
The rate at which patients received thrombolytic therapy, cardiac catheterization, or bypass surgery changed at similar rates in high- and low-impact groups among both insured and uninsured patients (results not shown).
Comparison of Effects in High- Versus Low-Impact Hospitals: Insured Patients
Our adjusted analyses of the change in each of these measures from before BBA to after BBA in high- versus low-impact hospitals provide information about the magnitude and statistical significance of these changes each year compared with baseline (Table 4). Among insured patients, time from hospital arrival to thrombolytic therapy ordering worsened in high- compared with low-impact hospitals from before BBA to 2000 (0.12 hours; P=0.03) and otherwise changed at similar rates in high- and low-impact hospitals. There was a similar pattern of results in time to thrombolytic therapy initiation in that the worsening in high- versus low-impact hospitals was significant from before BBA to 2000 only (P=0.01). Time from hospital arrival to arrival in the cardiac catheterization laboratory showed a significant improvement in high-impact hospitals relative to low-impact hospitals from before BBA to 2001 (–0.55 hours; P=0.05). Time to balloon inflation changed at similar rates in high- and low-impact hospitals. There were no differential changes in the rate at which patients received either thrombolytic therapy or PTCA in high- versus low-impact hospitals.
Medication use within 24 hours changed at similar rates among the insured in high- versus low-impact hospitals for aspirin, -blockers, and ACE inhibitors. There was a relative decrease in appropriate medication use at discharge among the insured for ACE inhibitors from before BBA to 1999 (odds ratio [OR], 0.85; P=0.01). There were no significant relative increases in the rate of appropriate medication use either within 24 hours of admission or on discharge in high- versus low-impact hospitals.
We found no evidence that mortality worsened to a greater degree in high- versus low-impact hospitals among the insured except for before BBA to 2001 (OR, 1.19; P=0.01; Table 5).
Comparison of Effects in High- Versus Low-Impact Hospitals: Uninsured Patients
Time from hospital arrival to thrombolytic therapy ordering worsened to a greater degree in high- versus low-impact hospitals among the uninsured only from before BBA to 1999 (0.30 hours; P=0.006), and time to initiation of thrombolytic therapy changed at similar rates in high- and low-impact hospitals. In contrast to the significant improvement in time from hospital arrival to arrival in catheterization laboratory in high- relative to low-impact hospitals among the insured, no significant differences were seen in the rate of change between high- and low-impact hospitals among the uninsured. Time to balloon inflation changed at similar rates in high- and low-impact hospitals.
Medication use within 24 hours changed at similar rates for aspirin, -blockers, and ACE inhibitors in high- and low-impact hospitals among the uninsured compared with the insured. There was a relative decrease in appropriate medication use at discharge among the uninsured for aspirin from before BBA to 2000 (OR, 0.54; P=0.05). There were no significant relative increases in the rate of appropriate medication use either within 24 hours of admission or on discharge in high- versus low-impact hospitals.
Mortality did not worsen to a greater degree in high- versus low-impact hospitals among the uninsured, and there was no significant difference in the degree to which mortality worsened in high- versus low-impact hospitals among the uninsured compared with the insured (Table 5).
Other Analyses
Inclusion of baseline operating margins to account for the financial starting position of hospitals before BBA did not change the sign or significance of the results, even among the uninsured. Results were also highly similar after adjustment for fund balances at baseline. There were no significant differences in the degree of change in length of stay between the high- and low-impact groups.
Discussion
Our results show that there was no consistent worsening in the process of AMI care in hospitals that were affected to a greater degree by the Medicare BBA. The rate of appropriate medication use generally changed at similar rates in high- and low-impact hospitals, and although time to thrombolytic therapy generally worsened in high-impact hospitals relative to low-impact hospitals, time to arrival in the cardiac catheterization laboratory generally improved in high-impact hospitals relative to low-impact hospitals. Mortality also changed at similar rates in the high- and low-impact groups except for before BBA to 2001 among the insured. There was no systematic pattern of disproportionate worsening of either process measures or outcomes among the uninsured.
There are 2 likely explanations for these results. The first is that operating margins declined at rates similar to the degree of reduction in net revenues, suggesting that hospitals cut their profit margins rather than shortchanging clinical care. This was particularly true in high-impact hospitals in which the reduction in net revenues of 3.2% to 4.7% was mirrored by a reduction in operating margins of 5.4% by 2000. In low-impact hospitals, net revenues declined by 1.5% to 2.2%, and operating margins declined 2.8% from baseline to 1999 before rebounding somewhat in 2000 and 2001. As a result, any differential reduction in net revenues between high- and low-impact hospitals may have been absorbed by the larger reduction in operating margins within high-impact hospitals. A second possible explanation for the lack of any systematic negative impact on AMI care of the BBA is that hospitals may prioritize AMI care because cardiovascular services tend to be high profile and profitable, so cuts in clinical services may be likely to target other areas. Even if decreasing margins led to skimping on care for the uninsured, the effects might more likely be seen in lower-profile or less profitable services.
In other work, we found no impact of the BBA on 30-day mortality for the following conditions: AMI, hip fracture, stroke, and gastrointestinal bleeding.20 We did find larger increases in high-impact hospitals in mortality among general surgery and orthopedic patients who experienced postoperative complications.21 We believe that the higher mortality rates observed among patients in financially stressed hospitals after surgery can be attributed to the fact that these patients are sicker and have much higher mortality rates at baseline, making any changes in the care process more likely to result in adverse outcomes. In addition, although hospital financial strain may have less impact on process of care for AMI, a condition for which extensive treatment guidelines exist and appropriate practices are well defined, it may be more likely that financial stress will affect quality of care in treating conditions for which there is more ambiguity as to the standard of care. It is important to emphasize that the findings we observed for patients with AMI may not reflect other areas of care provision within hospitals that may be more sensitive to cuts in reimbursement.
There is also evidence from other policy contexts that cuts in reimbursement can affect quality of care. Reductions in Medicaid reimbursements in California in the 1980s affected the number of services per admission to a greater degree in Medicaid-dependent hospitals than in hospitals with a smaller proportion of Medicaid patients, and services were reduced more significantly for Medicaid than privately insured patients.22 Reductions in Medicare PPS reimbursement from 1985 to 1994 increased 30-day mortality of AMI patients.23
Several data considerations should be kept in mind when our results are interpreted. First, the data are limited to patients with AMI, and we do not know whether these results generalize to other conditions. Second, hospitals that participate consistently in the NRMI registry may be less likely to reduce clinical service provision for AMI in the face of financial stress than other hospitals. Although we were careful to adjust for an extensive set of patient and hospital characteristics, changes in quality of care measured in high- versus low-impact hospitals may be attributable to other factors that are correlated with both impact group and outcomes. We did not adjust for multiple comparisons, although this limitation is probably less important in this context because the BBA was found to have little effect on either process or outcome measures.
Hospital financial stress, if severe enough, must at some point affect the type and quality of clinical service delivery. Hospitals that face reductions in net revenues can either reduce operating margins, preserving the amount of money that goes into patient care, or reduce operating expenses (the amount of money going into patient care) to preserve their operating margins. Although the BBA had a moderate impact on hospital revenues, it appears that hospitals within the NRMI sample chose to buffer the patient care process from these revenue reductions through reduced operating margins. Perhaps as a result, in contrast to the effects observed under some other policy reforms, care for the uninsured was not disproportionately affected.
In the long run, reductions in margins may translate into reductions in capital investments, infrastructure improvements, and quality improvement initiatives that could reduce the rate of improvement in care. However, in the short term, reductions in operating margins to the degree observed in this study may be a viable approach for hospitals to preserve the quality and type of services provided in the face of revenue reductions.
Acknowledgments
We thank VA HSR&D and the Doris Duke Charitable Foundation for funding support. Dr Volpp has received a VA HSR&D Career Development Award and a Doris Duke Foundation Clinical Scientist Development Award.
References
Rivers PA, Tsai K-L. The impact of the Balanced Budget Act of 1997 on Medicare in the USA: the fallout continues. Int J Health Care Qual Assur Inc Leadersh Health Serv. 2002; 15: 249–254.
American Hospital Association. Testimony of the American Hospital Association before the Subcommittee on Health and Environment of the Committee on Commerce of the US House of Representatives on Balanced Budget Act of 1997: impact on cost savings and patient care: September 15 1999. Available at: http://www.hospitalconnect.com/aha/advocacy-grassroots/advocacy/testimony/1999/testimony91599.html. Accessed July 29, 2003.
American Hospital Association. Statement of the American Hospital Association before the Commerce Subcommittee on Health and Environment of the United States House of Representatives on the impact of the Balanced Budget Act of 1997 on providers and patients, July 19 2000. Available at: http://www.hospitalconnect.com/aha/advocacy-grassroots/advocacy/testimony/2000/BBARelief70019.html. Accessed July 29, 2003.
Volpp, Kevin GM, Waldfogel J, Williams SV, Silber JH, Schwartz, JS, Pauly, MV. The effect of the New Jersey Health Care Reform Act on mortality from acute myocardial infarction. Health Serv Res. 2003; 38: 515–533.
Gruber J. The effect of competitive pressure on charity: hospital responses to price shopping in California. J Health Econ. 1994; 38: 183–212.
Kohn L, Corrigan J, Donaldson M. To Err Is Human: Building a Safer Health System. Washington, DC: National Academy Press; 2000.
Committee on Quality of Health Care in America, Institute of Medicine. Crossing the Quality Chasm. Washington, DC: National Academy Press; 2001.
Antman EM, Anbe DT, Armstrong PW, Bates ER, Green LA, Hand M, Hochman JS, Krumholz HM, Kushner FG, Lamas GA, Mullany CJ, Ornato JP, Pearle DL, Sloan MA, Smith SC Jr, Alpert JS, Anderson JL, Faxon DP, Fuster V, Gibbons RJ, Gregoratos G, Halperin JL, Hiratzka LF, Hunt SA, Jacobs AK; American College of Cardiology; American Heart Association Task Force on Practice Guidelines. ACC/AHA guidelines for the management of patients with ST-elevation myocardial infarction: executive summary: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Writing Committee to Revise the 1999 Guidelines for the Management of Patients With Acute Myocardial Infarction). Circulation. 2004; 110: 588–636.
American Hospital Association. BBA impact worksheet. Available at: http://www.aha.org/. Accessed June 24, 2002.
Carter GM, Newhouse JP, Relles DA. How much change in the case mix index is DRG creep J Health Econ. 1990; 9: 411–428.
Keeler EB, Rubenstein LV, Kahn KL, Draper D, Harrison ER, McGinty MJ, Rogers WH, Brook RH. Hospital characteristics and quality of care. JAMA. 1992; 268: 1709–1714.
Aiken LH, Clarke SP, Sloane DM, Sochalski J, Silber JH. Hospital nurse staffing and patient mortality, nurse burnout, and job dissatisfaction. JAMA. 2002; 288: 1987–1993.
Kuhn EM, Hartz AJ, Gottlieb MS, Rimm AA. The relationship of hospital characteristics and the results of peer review in six large states. Med Care. 1991; 29: 1028–1038.
Sheikh K, Bullock C. Urban-rural differences in the quality of care for Medicare patients with acute myocardial infarction. Arch Intern Med. 2001; 161: 737–743.
Taylor DH, Whellan DJ, Sloan FA. Effects of admission to a teaching hospital on the cost and quality of care for Medicare beneficiaries. N Engl J Med. 1999; 340: 293–299.
Ayanian JZ, Weissman JS. Teaching hospital and quality of care: a review of the literature. Milbank Q. 2002; 80: 569–593.
Newhouse JP. Toward a theory of nonprofit institutions: an economic model of the hospital. Am Econ Rev. 1970; 60: 64–74.
White H. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica. 1980; 48: 817–838.
Konetzka RT, Zhu J, Volpp K. Did the Balanced Budget Act hit teaching hospitals harder Acad Med. In press.
Seshamani M, Volpp KG. The effect of cuts in Medicare reimbursement on quality of hospital care. Presented at: Academy Health Annual Health Services Research Meeting; June 2004; San Diego, Calif.
Seshamani M, Zhu J, Volpp KG. The effect of cuts in Medicare reimbursement on postoperative mortality. Presented at: American Economic Association National Meeting; January 2005; Philadelphia, Penn.
Dranove D, White WD. Medicaid-dependent hospitals and their patients: how have they fared Health Serv Res. 1998; 33: 163–185.
Shen Y-C. The effect of financial pressure on the quality of care in hospitals. J Health Econ. 2003; 833: 1–27.(Kevin G. Volpp, MD, PhD; )
University of Chicago (R.T.K.), Chicago, Ill
Ovation Research Group (L.P.), Seattle, Wash
Duke University (E.P.), Durham, NC.
Abstract
Background— The Balanced Budget Act (BBA) of 1997 was designed to reduce Medicare reimbursements by $116.4 billion from 1998 to 2002. The objective of this study was to determine whether the process of care for acute myocardial infarction (AMI) worsened to a greater degree in hospitals under increased financial strain from the BBA and whether vulnerable populations such as the uninsured were disproportionately affected.
Methods and Results— We examined how process-of-care measures and in-hospital mortality for AMI patients changed in accordance with the degree of BBA-induced financial stress using data on 236 506 patients from the National Registry of Myocardial Infarction (NRMI) and Medicare Cost Reports from 1996 to 2001. BBA-induced reductions in hospital net revenues were estimated at 1.5% ($2.9 million) for hospitals with low BBA impact and 3.2% ($3.7 million) for hospitals with a high impact in 1998, worsening to 2.2% ($4.4 million) and 4.7% ($6.0 million), respectively, by 2001. For both insured and uninsured patients in high- versus low-impact hospitals, there was no systematic worsening of time to thrombolytic therapy, balloon inflation, medication use on admission, medication use on discharge, or mortality. There was no systematic pattern of different treatment among the insured and uninsured. Operating margins decreased to a degree commensurate with the degree of revenue reduction in high- versus low-impact hospitals.
Conclusions— BBA created a moderate financial strain on hospitals. However, process-of-care measures for both insured and uninsured patients with AMI were not appreciably affected by these revenue reductions. It is important to note that these results apply only to AMI patients; we do not know the degree to which these findings generalize to other conditions.
Key Words: delivery of health care health policy outcome assessment (health care)
Introduction
The Balanced Budget Act (BBA) of 1997 was designed to reduce Medicare reimbursements to hospitals, physicians, home health agencies, and skilled nursing facilities by $116.4 billion from 1998 to 2002.1 As a result of significant protests from teaching hospitals and the American Hospital Association, which cited "service closures and cutbacks,"2,3 Congress passed 2 laws to restore a portion of these cuts: the Balanced Budget Refinement Act (BBRA) in 1999, which was expected to restore $8.4 billion, and the Benefits and Improvement Protection Act (BIPA) in 2000, which was expected to restore a further $11.5 billion of the original $116 billion reductions.
The impact of reimbursement policy changes on quality of care has been seen in state-level analyses. The uninsured, whose care is funded largely from hospital margins, appears to be the most vulnerable group. Price-competitive market reforms in New Jersey in the early 1990s were associated with rises in inpatient mortality rates for uninsured but not insured acute myocardial infarction (AMI) patients and with concomitant decreases in the rates of revascularization and cardiac catheterization.4 Similarly, the enactment of legislation in California to permit selective contracting for insurance companies (thereby allowing them to price-shop among hospitals) decreased the quantity of services provided to the uninsured.5
We focus on AMI patients because cardiac care tends to represent a meaningful proportion of hospital budgets and patients and therefore is likely to be affected by financial pressures. In contrast to many other conditions, AMI patients are always admitted to hospitals, mitigating the importance of selection bias resulting from varying admission thresholds in studies of changes in hospital quality over time. Numerous studies have examined the effects of policies on AMI mortality for this reason, but few have examined process measures of quality, which may be more sensitive to changes in hospital finances than mortality and thus more reflective of factors under the hospital’s control. Detailed process measures of quality are available for AMI care but not for other conditions. It has become increasingly clear that systems factors have a great influence on quality of care,6,7 and such systems can be expensive to design and implement. The rate of provision of high-quality systems to ensure proper provision of medications and timely delivery of cardiac interventions may thus be affected by hospital financial stress.
We used data from the National Registry of Myocardial Infarction (NRMI) from 1996 to 2001 to examine whether quality of care worsened to a greater degree over time in hospitals under more financial strain from the BBA. We examined a set of measures known to be markers of quality of care for AMI patients8: in-hospital mortality; time to treatment with either primary angioplasty or thrombolytic therapy for patients with ST-segment elevation AMI; and rates of use of aspirin, -blockers, and ACE inhibitors within 24 hours of admission and on discharge. We focused on whether outcomes for the uninsured worsened to a greater degree than for the insured because cost-saving reforms such as the BBA could be a mechanism contributing to larger disparities in health outcomes.
Methods
This project was approved by the University of Pennsylvania Institutional Review Board.
Data
The NRMI is a database sponsored by Genentech that collects detailed data on process and outcomes of care of AMI patients from the approximately one third of all hospitals throughout the United States that volunteer to participate. NRMI includes data on hospitals from every state in the country. These data include information on time from symptom onset to arrival in the hospital, receipt of medications within 24 hours of admission and on discharge, time from hospital arrival to ordering and initiation of thrombolytic therapy, and discharge status. We limited our sample to hospitals that contributed at least 20 AMI cases per year to the NRMI registry from 1996 to 2001 because observed changes in process may not be meaningful when very few cases are addressed. We used data from the Medicare Cost Reports for information on net revenues and hospital operating margins and data from the American Hospital Association (AHA) Annual Survey for information on hospital characteristics.
Data Sample
Our initial sample included all interventional hospitals (hospitals with on-site angioplasty [PTCA] capabilities) that participated in the NRMI. The initial data set contained 567 546 admissions for AMI from 1996 to 2001 in 368 hospitals. We excluded hospitals with missing AHA data or unreliable or missing Medicare Cost Report information (88 hospitals; n=145 966); patients with unknown race (n=2149); patients who were transferred to other hospitals, because their ultimate discharge disposition was unknown (n=16 782); and patients transferring from other hospitals, because our focus was on "de novo" cases of AMI (n=125 307). After the above exclusions, hospitals that no longer had at least 20 admissions per year were excluded (72 hospitals; n=40 836). The final data set had 236 506 patients from 208 hospitals; 95 545 patients from 208 hospitals were admitted with ST-segment elevation AMI or left bundle-branch block, and time to thrombolytic therapy and time to balloon inflation were measured only for these patients. The sample of excluded patients did not differ systematically from the included sample. Because differential rates of change in transfer rates in high- and low-impact hospitals could bias the sample, we examined whether transfer rates changed at similar rates for high- and low-impact hospitals from before BBA (1996–1997) to after BBA (1998–2001). Among all hospitals that admitted at least 20 patients per year, transfer rates increased from 28.8% before BBA to 30.0% after BBA among hospitals in the high-impact quartile compared with an increase from 29.0% to 30.2% in the low-impact quartile. The degree of change was clearly quite similar between these 2 groups.
Calculation of the BBA Impact
We used a simulator constructed by the AHA as a tool for hospitals to calculate the financial impact of BBA and BBRA.9 We did not include BIPA in this simulation because these changes were instituted starting in 2001 and had a small financial impact in that year. Using financial information from the 1997 Medicare Cost Reports, the worksheet applies trend factors to simulate hospital Medicare reimbursements without the BBA (ie, under inflation), under the BBA, and under the BBRA. Subtracting the without-BBA estimations from the BBA and BBRA estimations provides a prediction of the financial impacts of the Medicare changes on reimbursements from fiscal 1998 to fiscal 2001 for each hospital. The worksheet includes most aspects of the BBA impact on hospitals, including diagnosis-related group payments, indirect medical education payments, and disproportionate share payments. Unlike measures such as operating margin and net revenue, it avoids contamination of the measure of financial impact by behavioral responses such as diagnosis-related group creep, whereby hospitals may respond to a change in financial reimbursement by altering reporting and charging,10 or adjustments in hospital payer mix. To obtain an estimate of the effect of the BBA on total hospital net revenues, we multiplied the percent change in Medicare reimbursements by the percentage of net patient revenue from Medicare reimbursements in the baseline year of 1997. We used an average BBA impact for each hospital over the 4 years of 1998 through 2001 to divide hospitals into high-impact hospitals (with the highest quartile of impact), low-impact hospitals (with the lowest quartile of BBA impact), and middle-impact hospitals (Table 1).
Adjustment for Patient and Hospital Characteristics
The NRMI includes extensive data on patient characteristics at admission that were used for risk adjustment: patient age, gender, and race; personal history of smoking, hypertension, hypercholesterolemia, congestive heart failure, stroke, angina, angioplasty, or CABG; family history of coronary artery disease; Killip congestive heart failure class; and time from symptom onset to arrival in the hospital. We also adjusted for blood pressure on admission using dummy variables for systolic blood pressure <60, 60 to 80, 81 to 100, and >100 mm Hg and for pulse on admission <50, 51 to 100, and >100 bpm.
On the basis of literature examining the link between hospital characteristics and quality of care,11–16 we adjusted for hospital characteristics that may influence health outcomes: bed size, teaching status (member of the Council of Teaching Hospitals), urban or rural setting, and availability of technology (onsite open heart surgery capabilities). We also examined whether nurse staffing ratios (ratio of RN and LPN to staffed bed) or operating margins changed to a greater degree in high- compared with low-impact hospitals and whether inclusion of these characteristics in our models affected measured changes in process or outcomes measures.
Statistical Analyses
We first examined the degree to which the BBA affected net revenues at different hospitals and divided hospitals into the top 25%, middle 50%, and bottom 25%. We then examined time trends in each study outcome among hospitals in the highest- and lowest-impact groups.
A patient-level logistic regression analysis was used to assess the effect of the BBA on the probability of in-hospital mortality; on receipt of aspirin, -blockers, or ACE inhibitors within 24 hours of admission or on discharge; and on receiving catheterization, angioplasty, or CABG. Linear regression was used to examine the effects of BBA on time to treatment for ST-segment elevation AMI patients. We adjusted for patient and hospital characteristics, intertemporal trends that were common to all hospitals in the sample, and baseline differences in each measure by BBA impact group. The effect of financial stress was identified by use of an interaction term between year and BBA impact group to compare how each study measure changed from before BBA to each of the post-BBA years in the high- compared with low-impact group.
We examined changes in each of our outcome measures separately for the insured and uninsured to test whether, within each of these groups, there were differential changes in the outcome measures for high- and low-impact hospitals. We then combined the data from the insured and uninsured to test whether effect sizes were different among the insured and uninsured. Because the ability of a hospital to respond to the financial stress resulting from Medicare cuts may depend on the amount of slack in financial resources at baseline, we examined the sensitivity of our results to inclusion of baseline operating margins in the regressions. We focused on baseline operating margin as a proxy for the amount of slack in financial resources because extensive economics literature indicates that hospitals trade off between operating margins and quality above the profit-maximizing level in the face of net revenue reductions.5,17 Given this, operating margins at baseline are probably the best proxy for the degree to which these tradeoffs have been made by hospitals before BBA. We further test whether adjusting for fund balances (assets minus liabilities) at baseline affected the quality response to BBA. Finally, we tested whether the BBA affected length of stay among the insured or uninsured or operating margins, examining each of these in turn as dependent variables.
Observations among patients from the same hospital may be correlated, and using aggregate variables such as hospital-wide financial impact in a patient-level regression can potentially lead to overestimation of the significance of such variables. To adjust for these problems, we used regression with robust standard errors clustered on the hospital level.18 For all of our models, we used generalized estimating equations that account for clustering of patients within hospitals. The statistical program used was SAS version 8.02. For both the linear and logistic regressions, the GENMOD procedure was used.
Results
Descriptive Analysis
Medication use within 24 hours of hospital admission and on discharge increased in both the high- and low-impact groups and among both the insured and uninsured (Table 3). Time to thrombolytic therapy ordering and initiation decreased slightly in the low-impact group among the insured (0.1 hours) while remaining constant or increasing slightly among the insured within the high-impact group and remaining constant or decreasing slightly among the uninsured in both impact groups. Door to balloon inflation times decreased more in high-impact hospitals among the insured (0.5 hours) than in low-impact hospitals (0.3 hours). The year-to-year trends for door to balloon time and door to initiation of thrombolytic therapy are illustrated in the Figure. This figure shows that the time to treatment is similar among the insured and uninsured and that there is no clear worsening in high-impact hospitals relative to low-impact hospitals except to a small degree in initiation of thrombolysis among the insured. Mortality declined slightly in both the low- and high-impact groups from before BBA to after BBA in unadjusted analyses except for an increase among the uninsured in the high-impact group.
Changes in time to treatment in high- vs low-impact hospitals over time.
Changes in Procedure Rates in High- Versus Low-Impact Hospitals
The rate at which patients received thrombolytic therapy, cardiac catheterization, or bypass surgery changed at similar rates in high- and low-impact groups among both insured and uninsured patients (results not shown).
Comparison of Effects in High- Versus Low-Impact Hospitals: Insured Patients
Our adjusted analyses of the change in each of these measures from before BBA to after BBA in high- versus low-impact hospitals provide information about the magnitude and statistical significance of these changes each year compared with baseline (Table 4). Among insured patients, time from hospital arrival to thrombolytic therapy ordering worsened in high- compared with low-impact hospitals from before BBA to 2000 (0.12 hours; P=0.03) and otherwise changed at similar rates in high- and low-impact hospitals. There was a similar pattern of results in time to thrombolytic therapy initiation in that the worsening in high- versus low-impact hospitals was significant from before BBA to 2000 only (P=0.01). Time from hospital arrival to arrival in the cardiac catheterization laboratory showed a significant improvement in high-impact hospitals relative to low-impact hospitals from before BBA to 2001 (–0.55 hours; P=0.05). Time to balloon inflation changed at similar rates in high- and low-impact hospitals. There were no differential changes in the rate at which patients received either thrombolytic therapy or PTCA in high- versus low-impact hospitals.
Medication use within 24 hours changed at similar rates among the insured in high- versus low-impact hospitals for aspirin, -blockers, and ACE inhibitors. There was a relative decrease in appropriate medication use at discharge among the insured for ACE inhibitors from before BBA to 1999 (odds ratio [OR], 0.85; P=0.01). There were no significant relative increases in the rate of appropriate medication use either within 24 hours of admission or on discharge in high- versus low-impact hospitals.
We found no evidence that mortality worsened to a greater degree in high- versus low-impact hospitals among the insured except for before BBA to 2001 (OR, 1.19; P=0.01; Table 5).
Comparison of Effects in High- Versus Low-Impact Hospitals: Uninsured Patients
Time from hospital arrival to thrombolytic therapy ordering worsened to a greater degree in high- versus low-impact hospitals among the uninsured only from before BBA to 1999 (0.30 hours; P=0.006), and time to initiation of thrombolytic therapy changed at similar rates in high- and low-impact hospitals. In contrast to the significant improvement in time from hospital arrival to arrival in catheterization laboratory in high- relative to low-impact hospitals among the insured, no significant differences were seen in the rate of change between high- and low-impact hospitals among the uninsured. Time to balloon inflation changed at similar rates in high- and low-impact hospitals.
Medication use within 24 hours changed at similar rates for aspirin, -blockers, and ACE inhibitors in high- and low-impact hospitals among the uninsured compared with the insured. There was a relative decrease in appropriate medication use at discharge among the uninsured for aspirin from before BBA to 2000 (OR, 0.54; P=0.05). There were no significant relative increases in the rate of appropriate medication use either within 24 hours of admission or on discharge in high- versus low-impact hospitals.
Mortality did not worsen to a greater degree in high- versus low-impact hospitals among the uninsured, and there was no significant difference in the degree to which mortality worsened in high- versus low-impact hospitals among the uninsured compared with the insured (Table 5).
Other Analyses
Inclusion of baseline operating margins to account for the financial starting position of hospitals before BBA did not change the sign or significance of the results, even among the uninsured. Results were also highly similar after adjustment for fund balances at baseline. There were no significant differences in the degree of change in length of stay between the high- and low-impact groups.
Discussion
Our results show that there was no consistent worsening in the process of AMI care in hospitals that were affected to a greater degree by the Medicare BBA. The rate of appropriate medication use generally changed at similar rates in high- and low-impact hospitals, and although time to thrombolytic therapy generally worsened in high-impact hospitals relative to low-impact hospitals, time to arrival in the cardiac catheterization laboratory generally improved in high-impact hospitals relative to low-impact hospitals. Mortality also changed at similar rates in the high- and low-impact groups except for before BBA to 2001 among the insured. There was no systematic pattern of disproportionate worsening of either process measures or outcomes among the uninsured.
There are 2 likely explanations for these results. The first is that operating margins declined at rates similar to the degree of reduction in net revenues, suggesting that hospitals cut their profit margins rather than shortchanging clinical care. This was particularly true in high-impact hospitals in which the reduction in net revenues of 3.2% to 4.7% was mirrored by a reduction in operating margins of 5.4% by 2000. In low-impact hospitals, net revenues declined by 1.5% to 2.2%, and operating margins declined 2.8% from baseline to 1999 before rebounding somewhat in 2000 and 2001. As a result, any differential reduction in net revenues between high- and low-impact hospitals may have been absorbed by the larger reduction in operating margins within high-impact hospitals. A second possible explanation for the lack of any systematic negative impact on AMI care of the BBA is that hospitals may prioritize AMI care because cardiovascular services tend to be high profile and profitable, so cuts in clinical services may be likely to target other areas. Even if decreasing margins led to skimping on care for the uninsured, the effects might more likely be seen in lower-profile or less profitable services.
In other work, we found no impact of the BBA on 30-day mortality for the following conditions: AMI, hip fracture, stroke, and gastrointestinal bleeding.20 We did find larger increases in high-impact hospitals in mortality among general surgery and orthopedic patients who experienced postoperative complications.21 We believe that the higher mortality rates observed among patients in financially stressed hospitals after surgery can be attributed to the fact that these patients are sicker and have much higher mortality rates at baseline, making any changes in the care process more likely to result in adverse outcomes. In addition, although hospital financial strain may have less impact on process of care for AMI, a condition for which extensive treatment guidelines exist and appropriate practices are well defined, it may be more likely that financial stress will affect quality of care in treating conditions for which there is more ambiguity as to the standard of care. It is important to emphasize that the findings we observed for patients with AMI may not reflect other areas of care provision within hospitals that may be more sensitive to cuts in reimbursement.
There is also evidence from other policy contexts that cuts in reimbursement can affect quality of care. Reductions in Medicaid reimbursements in California in the 1980s affected the number of services per admission to a greater degree in Medicaid-dependent hospitals than in hospitals with a smaller proportion of Medicaid patients, and services were reduced more significantly for Medicaid than privately insured patients.22 Reductions in Medicare PPS reimbursement from 1985 to 1994 increased 30-day mortality of AMI patients.23
Several data considerations should be kept in mind when our results are interpreted. First, the data are limited to patients with AMI, and we do not know whether these results generalize to other conditions. Second, hospitals that participate consistently in the NRMI registry may be less likely to reduce clinical service provision for AMI in the face of financial stress than other hospitals. Although we were careful to adjust for an extensive set of patient and hospital characteristics, changes in quality of care measured in high- versus low-impact hospitals may be attributable to other factors that are correlated with both impact group and outcomes. We did not adjust for multiple comparisons, although this limitation is probably less important in this context because the BBA was found to have little effect on either process or outcome measures.
Hospital financial stress, if severe enough, must at some point affect the type and quality of clinical service delivery. Hospitals that face reductions in net revenues can either reduce operating margins, preserving the amount of money that goes into patient care, or reduce operating expenses (the amount of money going into patient care) to preserve their operating margins. Although the BBA had a moderate impact on hospital revenues, it appears that hospitals within the NRMI sample chose to buffer the patient care process from these revenue reductions through reduced operating margins. Perhaps as a result, in contrast to the effects observed under some other policy reforms, care for the uninsured was not disproportionately affected.
In the long run, reductions in margins may translate into reductions in capital investments, infrastructure improvements, and quality improvement initiatives that could reduce the rate of improvement in care. However, in the short term, reductions in operating margins to the degree observed in this study may be a viable approach for hospitals to preserve the quality and type of services provided in the face of revenue reductions.
Acknowledgments
We thank VA HSR&D and the Doris Duke Charitable Foundation for funding support. Dr Volpp has received a VA HSR&D Career Development Award and a Doris Duke Foundation Clinical Scientist Development Award.
References
Rivers PA, Tsai K-L. The impact of the Balanced Budget Act of 1997 on Medicare in the USA: the fallout continues. Int J Health Care Qual Assur Inc Leadersh Health Serv. 2002; 15: 249–254.
American Hospital Association. Testimony of the American Hospital Association before the Subcommittee on Health and Environment of the Committee on Commerce of the US House of Representatives on Balanced Budget Act of 1997: impact on cost savings and patient care: September 15 1999. Available at: http://www.hospitalconnect.com/aha/advocacy-grassroots/advocacy/testimony/1999/testimony91599.html. Accessed July 29, 2003.
American Hospital Association. Statement of the American Hospital Association before the Commerce Subcommittee on Health and Environment of the United States House of Representatives on the impact of the Balanced Budget Act of 1997 on providers and patients, July 19 2000. Available at: http://www.hospitalconnect.com/aha/advocacy-grassroots/advocacy/testimony/2000/BBARelief70019.html. Accessed July 29, 2003.
Volpp, Kevin GM, Waldfogel J, Williams SV, Silber JH, Schwartz, JS, Pauly, MV. The effect of the New Jersey Health Care Reform Act on mortality from acute myocardial infarction. Health Serv Res. 2003; 38: 515–533.
Gruber J. The effect of competitive pressure on charity: hospital responses to price shopping in California. J Health Econ. 1994; 38: 183–212.
Kohn L, Corrigan J, Donaldson M. To Err Is Human: Building a Safer Health System. Washington, DC: National Academy Press; 2000.
Committee on Quality of Health Care in America, Institute of Medicine. Crossing the Quality Chasm. Washington, DC: National Academy Press; 2001.
Antman EM, Anbe DT, Armstrong PW, Bates ER, Green LA, Hand M, Hochman JS, Krumholz HM, Kushner FG, Lamas GA, Mullany CJ, Ornato JP, Pearle DL, Sloan MA, Smith SC Jr, Alpert JS, Anderson JL, Faxon DP, Fuster V, Gibbons RJ, Gregoratos G, Halperin JL, Hiratzka LF, Hunt SA, Jacobs AK; American College of Cardiology; American Heart Association Task Force on Practice Guidelines. ACC/AHA guidelines for the management of patients with ST-elevation myocardial infarction: executive summary: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Writing Committee to Revise the 1999 Guidelines for the Management of Patients With Acute Myocardial Infarction). Circulation. 2004; 110: 588–636.
American Hospital Association. BBA impact worksheet. Available at: http://www.aha.org/. Accessed June 24, 2002.
Carter GM, Newhouse JP, Relles DA. How much change in the case mix index is DRG creep J Health Econ. 1990; 9: 411–428.
Keeler EB, Rubenstein LV, Kahn KL, Draper D, Harrison ER, McGinty MJ, Rogers WH, Brook RH. Hospital characteristics and quality of care. JAMA. 1992; 268: 1709–1714.
Aiken LH, Clarke SP, Sloane DM, Sochalski J, Silber JH. Hospital nurse staffing and patient mortality, nurse burnout, and job dissatisfaction. JAMA. 2002; 288: 1987–1993.
Kuhn EM, Hartz AJ, Gottlieb MS, Rimm AA. The relationship of hospital characteristics and the results of peer review in six large states. Med Care. 1991; 29: 1028–1038.
Sheikh K, Bullock C. Urban-rural differences in the quality of care for Medicare patients with acute myocardial infarction. Arch Intern Med. 2001; 161: 737–743.
Taylor DH, Whellan DJ, Sloan FA. Effects of admission to a teaching hospital on the cost and quality of care for Medicare beneficiaries. N Engl J Med. 1999; 340: 293–299.
Ayanian JZ, Weissman JS. Teaching hospital and quality of care: a review of the literature. Milbank Q. 2002; 80: 569–593.
Newhouse JP. Toward a theory of nonprofit institutions: an economic model of the hospital. Am Econ Rev. 1970; 60: 64–74.
White H. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica. 1980; 48: 817–838.
Konetzka RT, Zhu J, Volpp K. Did the Balanced Budget Act hit teaching hospitals harder Acad Med. In press.
Seshamani M, Volpp KG. The effect of cuts in Medicare reimbursement on quality of hospital care. Presented at: Academy Health Annual Health Services Research Meeting; June 2004; San Diego, Calif.
Seshamani M, Zhu J, Volpp KG. The effect of cuts in Medicare reimbursement on postoperative mortality. Presented at: American Economic Association National Meeting; January 2005; Philadelphia, Penn.
Dranove D, White WD. Medicaid-dependent hospitals and their patients: how have they fared Health Serv Res. 1998; 33: 163–185.
Shen Y-C. The effect of financial pressure on the quality of care in hospitals. J Health Econ. 2003; 833: 1–27.(Kevin G. Volpp, MD, PhD; )