|
MARKETWATCHMaking Profits And Providing Care: Comparing Nonprofit, For-Profit, And Government Hospitals
Three types of entities—nonprofit, for-profit, and government—own hospitals. Yet we know neither whether hospital types specialize in different medical services nor how service profitability affects specialization. In this econometric analysis of American Hospital Association data for every U.S. urban, acute care hospital (1988–2000), more than thirty services were categorized as relatively profitable, unprofitable, or variable. For-profits are most likely to offer relatively profitable medical services; government hospitals are most likely to offer relatively unprofitable services; nonprofits often fall in the middle. For-profits are also more responsive to changes in service profitability than the other two types.
How do for-profit hospitals make profits? One theory is that they manage their case-mix more carefully than nonprofit and government hospitals do. For example, they differentially locate in areas with relatively well-insured patients.1 Underlying this concept of profit making is the assumption that all general hospitals, regardless of ownership, are alike in the types of medical services they provide. They merely differ in their patient mix. This paper advances a different mechanism of profit making. It proposes that for-profit hospitals are more likely than other types to decide which medical services to offer based on service profitability. Under this theory, general hospitals do not provide a standard array of medical services. Instead, they specialize. Analyzing 1988–2000 data on medical service provision for every U.S. urban, general hospital, this paper tests whether hospital types specialize in services based on profitability. Which profit-making method hospitals adopt, and to what degree hospital types employ them, matters a great deal for tax policy, structuring reimbursement, and quality of care. Two-thirds of all U.S. urban hospitals are nonprofit, with the remainder split between for-profit and government ownership. These hospital types operate under different legal rules. For-profits may distribute accounting profits to shareholders, whereas government and nonprofit hospitals enjoy income and property tax exemptions. However, there is reason to expect all hospitals to provide a similar array of medical services: General hospitals all treat patients with a mix of needs, contract with the same insurers and government payers, operate under the same health regulations, and employ staff with the same training and ethical obligations. Consequently, it is not surprising that much of the empirical literature on corporate ownership finds little difference among hospital types.2 Although the ownership literature is voluminous and widely debated in the United States and abroad, it is incomplete.3 Studies have mainly examined financial topics such as costs, profits, billing, the value of uncompensated care, and responsiveness to financial pressure. Important subjects such as quality, physician control, and patient access have been studied less frequently, have focused on a limited number of services, and have generated inconclusive results.4 Researchers have not previously looked broadly at whether any hospital types systematically offer relatively profitable services and avoid relatively unprofitable ones. To that end, this study investigates how ownership affects the central activity of hospitals: medical care provision. It evaluates more than thirty services, ranging from sports medicine to AIDS treatment, to ask whether ownership is correlated with offering services and how those choices relate to profit seeking.
Hospital population. Data are from the American Hospital Association’s (AHA’s) Annual Surveys of Hospitals during 1988–2000 and the 1990 U.S. census. The AHA survey had almost a 20 percent nonresponse rate, and nonrespondents were disproportionately for-profit hospitals (1988: 3 percent nonprofit, 4 percent government, 18 percent for-profit; 2000: 14 percent nonprofit, 20 percent government, 26 percent for-profit). Since the same hospitals were not missing in all years, however, the results should be valid. The analysis included all nonrural, acute care hospitals that operate in metropolitan statistical areas (MSAs) with at least two general medical and surgical hospitals, excluding military, uncategorized federal, and prison hospitals. I excluded rural hospitals because there are relatively few for-profit hospitals in rural areas and they provide a limited range of services. In 1995, for example, of the roughly 2,500 rural hospitals, approximately 8 percent were for-profit, and only a few provided open-heart surgery.
Study variables.
Hospitals were classified by ownership: nonprofit, for-profit, and government. They were asked whether each service was offered at the hospital or another hospital, or was not available (1988–1993), or whether the service was offered at the hospital or a subsidiary, another location in the system, the network, or a joint venture (1994–2000). Because the study assessed hospital-based services, the first categories in the two survey versions were treated as equivalent. Roughly forty variables were included in the data on acute care services, of which thirty-two were sufficiently well-defined and consistent across the relevant time period to be useable (Exhibit 1
I imputed missing values using data from the years before and after the missing year. For the end years (1988, 2000) I based the imputation on the next or previous two years. Where several values were missing, I excluded the observations. Generally, I imputed fewer than 4 percent of observations for each service.
Determining profitability.
To test hospitals’ responsiveness to service profitability, I sorted services into three categories: relatively high, low, and variable profitability (see Exhibit 1 Because this project is concerned primarily with hospitals’ behavior and motivation, I checked the scientific literature with a review of trade publications, business magazines, and newspaper reports. The qualitative evaluation of relative service profitability was critical to the project because perceptions of whether a service would be profitable were likely as important determinants of service investment choices as whether services turned out to be profitable. The various sources yielded remarkably similar results.6 Abbreviated examples of three services follow. Open-heart surgery, for example, is relatively profitable. Like almost all surgical and related services, cardiac services—including cardiac catheterization labs, angioplasty, and coronary artery bypass graft (CABG)—are widely known to be hospital profit centers.7 Insurers typically reimburse heart attack treatments at high rates, patients receiving CABG are unusually well insured since most are covered by Medicare, and there is high and increasing spending on cardiac care.8 From 1984 to 1994, the real price for bypass surgery among Medicare patients rose 2.3 percent annually, from $29,176 to $36,564 (1991 dollars), while the share of patients receiving the treatment increased by one percentage point annually, from 5 percent to 15 percent.9 The costs of supplying CABG in real terms either were flat or fell during the period.10 In 1991, because spending on bypass surgery was so high, the Centers for Medicare and Medicaid Services (CMS, then HCFA) ran a pilot program in which hospitals and physicians negotiated prices.11 During the study period, for-profit corporations opened single-service cardiac surgery centers, while neighboring hospitals complained of losing profitable business.12 Conversely, hospital-based, psychiatric emergency services are relatively unprofitable for several reasons: (1) the emergency department (ED) is a comparatively unprofitable setting, which attracts patients whose admissions are more costly than those of patients who are admitted to the hospital by other means; (2) psychiatric care reimbursement is uncertain and often low relative to cost; and (3) the care attracts a poor, poorly insured, sick, and difficult-to-manage population.13 Compared with that of acute care, the profitability of postacute services varied dramatically during the 1980s and 1990s. With legal challenges to reimbursement resolved and the hospital prospective payment system (PPS) implemented, postacute services became highly profitable by the early 1990s.14 Unlike acute services diagnosis-related group (DRG) payments in which hospitals receive a per episode payment for each patient, Medicare paid hospitals a cost-related reimbursement for postacute services. Hospitals could increase reimbursements by unbundling the services and transferring patients to postacute care at the end of their hospital stay.15 There is considerable evidence of these transfers. From 1981 to 1998 the number of inpatient days for Medicare beneficiaries fell at an average annual rate of 4.1 percent; from 1986 to 1998 home health visits for Medicare beneficiaries grew at an average annual rate of 15.6 percent.16 Payments were particularly generous to new entrants, with skilled nursing facilities (SNFs) and home health services exempt from cost limits for the first years of operation. In fact, home health payments grew from $3.9 billion to more than $18.3 billion between 1990 and 1996.17 Hospital administrators and regulators alike recognized the profitability potential of postacute care.18 With passage of the 1997 Balanced Budget Act (BBA) of 1997, however, the profit-making opportunities plummeted. Medicare payments were reduced, the CMS developed a PPS for postacute services, and home health care spending fell by a factor of two.
Statistical analysis.
Using a probit model, I analyzed whether ownership was correlated with the probability of offering each of approximately thirty medical services (Exhibit 1 To correct for potential biases, I controlled for hospital and market characteristics. Hospital variables included size (measured by admissions), teaching status (measured by teaching association membership), and a dummy variable for location by geographic region. Market characteristics included patients’ sex, race, household income, and age. These were compiled using the ten-mile radii around the centers of the hospitals’ ZIP codes, the mean distance that captures 75 percent of discharges among acute care urban hospitals.19 I adjusted for heteroskedasticity, and, because the probability of offering a service is not independent among years, I allowed for an arbitrary covariance matrix within each hospital over time. By varying the hospital’s ownership type while holding the independent variables constant (at 1994 levels or the next-closest year), I predicted the probabilities that each hospital in each year would offer a service. I then averaged the individual predicted probabilities to obtain a single probability that a hospital type would offer a service each year. I performed sensitivity tests on the three services described above. Because size is the best predictor of service offering, I restricted the regressions to the observations in the top, bottom, and middle two quartiles of hospitals measured by admissions. To ensure that ownership, rather than geography, explained these results, I ran several sensitivity tests related to hospital regions. John Wennberg and others have observed that medical service provision varies considerably by small region.20 Consequently, one might think that firm types chose where to operate based on the local character of demand. I tested this alternative explanation for the results by using a fixed-effects approach, including an indicator variable for the year 2000 Hospital Referral Regions (HRRs) in which each hospital operates. I also altered the region variable to account for areas of high for-profit penetration (for example, the South and Southwest) and included dummy variables for all nine AHA regions. To test the sensitivity to other market characteristics, I added age-squared categories for the percentage of the population over age sixty-five and over age eighty. Because state payment policies for mental health services vary considerably, I included state dummies and state-year interactions for the psychiatric emergency service estimations. In addition, to test variation within the government hospital category, I excluded hospitals in the Department of Veterans Affairs (VA) system. Sensitivity tests also included propensity score analysis, a method used to make causal inferences when assignment to a group, such as nonprofit ownership status, is not random.21 The results confirmed that I compared hospitals that differed primarily by ownership and not other hospital characteristics such as hospital size. More specifically, I determined the conditional probability of corporate ownership (nonprofit versus for-profit; nonprofit versus government, government versus for-profit), given the observed characteristics used in the Probit estimates (the propensity scores), created five subcategories defined by the estimated propensity score, and predicted the probability of a hospital type’s offering a service in a given year, controlling for the propensity grouping.
The results demonstrate that among comparable hospitals, for-profits are more likely than nonprofits, which in turn are more likely than government hospitals, to offer open-heart surgery. The magnitude of these differences seems large, given the importance of the decision to offer the service and the constraints to doing so. For-profits are, on average, 13.0 percentage points more likely than government hospitals (40.9 percent versus 27.9 percent, p < .001) and 7.3 percentage points more likely than nonprofit hospitals (40.9 percent versus 33.6 percent, p <.001) to offer open-heart surgery (Exhibit 2  ). Restricting the data to the smallest hospitals—those in the bottom two admissions quartiles—the nonprofit-government difference for open-heart surgery was insignificant. This result is expected because so few small hospitals offer open-heart surgery.
Unlike open-heart surgery, for-profits are less likely than nonprofits, which in turn are less likely than government hospitals, to offer psychiatric emergency care (Exhibit 3 ). On average from 1988 to 2000, controlling for other characteristics, 41 percent of for-profit hospitals offered psychiatric emergency services, compared with 48 percent of nonprofit and 56 percent of government hospitals. Again, the magnitude of these differences is large: For-profits are 15.0 percentage points less likely than government hospitals and 8.4 percentage points less likely than nonprofit hospitals (p <.01) to offer psychiatric emergency care. Among the smallest hospitals, the nonprofit versus for-profit difference was insignificant, as was the difference between nonprofit and non-VA government hospitals.
Not only did the probability of offering home health care vary by ownership, but the relative differences among types varied over time. The probability of offering home health services when the service was profitable increased for all three hospital types. However, among for-profits, growth of home health care when profitable and decline when unprofitable were particularly dramatic (Exhibit 4 ). From 1988 to 1996, the probability of a for-profit hospital’s offering home health services more than tripled (17.5 percent to 60.9 percent), controlling for hospital and market characteristics. During the same period, the probability of offering home health care only grew slightly more than ten percentage points (40.9 percent to 51.7 percent) for nonprofit and fourteen percentage points (38.1 percent to 51.9 percent) for government hospitals. From 1997 to 2000, as home health care became relatively unprofitable with the implementation of the BBA of 1997, the probability of offering it fell a striking 37.5 percentage points among for-profit, 7.7 percentage points among nonprofit, and 1.5 percentage points among government hospitals.
Other services. Tests of more than thirty other services yielded similar results (Exhibit 5 ). While for-profit hospitals were only somewhat more likely than nonprofits to offer relatively profitable services, both for-profit and nonprofit hospitals were considerably more likely than government hospitals to offer relatively profitable services. For-profits were less likely than nonprofits, which in turn were less likely than government hospitals, to offer relatively unprofitable services. For-profit hospitals were more responsive than the other types were to rapid changes in service profitability (Exhibit 6). The average probability of a hospital type’s offering a medical service from 1988 through 2000, controlling for other characteristics, is reported in Exhibit 7.
Medical service offerings vary markedly by ownership, likely because hospital types adopt or prioritize goals differently. Although all hospitals must earn sufficient profits to operate, the evidence here suggests that for-profits are more likely to respond to profitability than the other types are when making supply decisions. Since government hospitals are most likely to supply the unprofitable services that are disproportionately needed by poor and underinsured patients, the evidence also suggests that such hospitals are caregivers of last resort. Nonprofit hospitals are often the intermediate type in terms of balancing profit seeking and serving the poor through service choices. Neither the profit-making versus charitable divide (for-profit versus nonprofit and government) nor the public versus private divide (government versus nonprofit and for-profit) fully predicts behavior. These results suggest that corporate form can be used as a regulatory tool when regulations are designed to account for the motivations and relative responsiveness to profitability of all three types. For example, ownership could be considered in designing reimbursement policies. In addition, if they hope to secure a full range of services, states’ attorneys general should consider the mix of types in a market when overseeing hospital conversions. A further implication is that the measure traditionally used to justify nonprofit tax exemptions—the provision of uncompensated care—is too limited. Recent calls to eliminate nonprofit subsidies, commonly founded on the claim that there are no important differences among corporate types, should be rejected.22 Differences in service mix matter to all patients, not only the uninsured and nonpaying patients. Although this paper has not addressed health outcomes, it does raise the question of the relationship between service profitability and medical appropriateness. As discussed in the home health example above, for-profits’ responsiveness to incentives is noteworthy for its magnitude and speed. However, we do not know whether the most medically appropriate mix of hospital services is the most profitable. Public payment rates are set through a complex and changing process based on, among other factors, the evolving judgment of rate setters, imperfect adjustments for hospital markets’ demographic and geographic characteristics, and the political strength of interested parties. Private payment rates result from complex negotiations and relative bargaining power. We need more study on how and, indeed, whether these processes produce incentives for hospitals to provide a medically appropriate service mix. Study limitations. Despite the methodological strength of this work—particularly the focus on medical services rather than financial behavior and the breadth of services studied—it has limitations. First, as discussed above, the nonrespondents were disproportionately for-profit. The data are self-reported and not independently verified, although there is no reason to suspect that data reliability is correlated with ownership. Second, despite the rigorous nature of the research used to determine profitability, a note of caution is necessary. Profitability is not an inherent attribute of medical services; rather, it depends on institution-specific factors such as management skills, case-mix, and local input costs. Further, even within a single hospital, costs and charges differ, discounts vary by individual payer, and allocation of joint costs blur the profitability picture. Despite these complications, however, one can reasonably compare the relative profitability of services defined as bluntly as they are in the AHA data. Whether nonprofit organizations behave differently from other ownership types, particularly in the hospital industry, has raised considerable debate. In 2004 more than fifty lawsuits alleging that nonprofit hospitals have violated their charitable obligations were filed in federal district courts alone.23 Congress is considering an extensive overhaul of the nonprofit sector’s regulatory regime to increase accountability.24 Any discussion of the value of nonprofit hospital ownership must account for the significant differences in service offerings among hospital types and how those offerings vary according to profitability.
Jill Horwitz (jrhorwit{at}umich.edu) is an assistant professor of law at the University of Michigan Law School in Ann Arbor. The author thanks Jia-Jia Ye for research assistance and Paula Payton for administrative assistance. She thanks Allan Brandt, Suzanne Cooper, David Cutler, Phoebe Ellsworth, Don Herzog, Nancy Keating, Mary Beth Landrum, Barbara McNeil, Martha Minow, Carl Morris, David Naylor, Joseph Newhouse, Edward Parson, Jonathan Skinner, Richard Zeckhauser, three anonymous referees, and participants in the November 2003 National Bureau of Economic Research (NBER) Health Care meetings for helpful comments. This study was supported by a grant from the National Institute on Aging to the NBER (Grant no. T32-AG00186); the Hauser Center for Nonprofit Organizations; the Harvard University doctoral program in health policy; and the University of Michigan Law School, John M. Olin Center for Law and Economics.
This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Top
Study Data And Methods
Study Results
