|
MARKETWATCHEvidence Of Cost Shifting In California Hospitals
We used 1993–2001 data from private hospitals in California to investigate whether decreases in Medicare and Medicaid prices were associated with increases in prices paid for privately insured patients. We found that a 1 percent relative decrease in the average Medicare price is associated with a 0.17 percent increase in the corresponding price paid by privately insured patients; similarly, a 1 percent relative reduction in the average Medicaid price is associated with a 0.04 percent increase. These relationships imply that cost shifting from Medicare and Medicaid to private payers accounted for 12.3 percent of the total increase in private payers’ prices from 1997 to 2001.
THERE IS CONTINUED controversy about whether hospitals increase prices to private payers in response to reductions in payment rates from public programs.1 The debate as to whether such cost shifting takes place is of increasing importance in policy discussions of how to reduce budget deficits. Recent history is suggestive but clearly not definitive in this regard. In the years following the enactment of the Balanced Budget Act (BBA) of 1997, Medicare and Medicaid payment rates to hospitals grew more slowly than costs, while payments by private insurers accelerated. Because slowing the growth of Medicare and Medicaid payment rates is likely to become an important part of any deficit reduction effort, it is important to know whether such a relative reduction in payment rates is related to an increase in private insurers’ payment rates.2 Economic theory suggests that two conditions must exist for hospital cost shifting to occur: Hospitals must be able to price above the level they would be at in highly competitive markets, and prices must not already be at profit-maximizing levels.3 One might expect cost shifting to occur because hospital markets are generally not very competitive and are dominated by not-for-profit hospitals; however, these conditions imply that cost shifting will be reduced, if not prevented, by increasingly intense price competition among hospitals.4 In a study of hospitals in California spanning 1983–1991—the period immediately following the enactment of legislation to encourage price competition in California’s health care system—we found that increases in prices to privately insured patients were strongly related to a reduced rate of growth in Medicare payments.5 However, it remained unclear whether this cost shifting reflected the fact that hospitals and managed care plans had not yet fully adjusted to the new competitive environment. This paper takes advantage of the confluence of two factors: (1) a dramatic increase in the intensity of price competition among California’s hospitals during the 1990s; and (2) the enactment of the BBA in 1997, which greatly reduced the rate of growth in Medicare reimbursement rates.6 We use California data for 1993–2001, spanning the years before and after implementation of the BBA, to examine the dynamics of cost shifting.
The primary data for these analyses were obtained from the Annual Disclosure Reports filed by each hospital in California with the Office of Statewide Health Planning and Development (OSHPD). Data were annualized to calendar years by combining relevant portions of adjacent fiscal years. We analyzed data for the 1993–2001 period; 2001 was the most recent complete calendar year with available data at the time of the analysis. The data included detailed utilization, cost, and revenue data by major payer categories (Medicare, Medicaid, and private payers). The primary advantages of these data are that (1) the data are disaggregated by payer across a comprehensive set of cost and revenue centers, which permits more accurate estimates of payer-specific prices and costs; and (2) improvements in data reporting now allow a more accurate identification of the utilization, costs, and revenues associated specifically with privately insured patients. Public hospitals were excluded from the analyses because their financing is fundamentally different from those of private for-profit and not-for-profit hospitals. The model we used has been described in detail elsewhere.7 It includes private insurers’ prices as the dependent variable and five independent variables: Medicare prices, Medicaid prices, average cost per adjusted day, a measure of market competition (the Hirschmann-Herfindahl Index, or HHI), and HHI interacted with the annual time dummies. We used a hospital fixed-effects model specification to estimate the relationships between changes in the dependent variable and corresponding changes in the independent variables. We included average costs as an independent variable because prices are likely to depend on costs, but because costs also could be endogenous, which partially reflects anticipated revenue, this variable was instrumented in a first-stage model.8 Extensive specification testing failed to detect any statistically significant difference in the cost-shifting relationships for for-profit and not-for-profit hospitals. Similarly, no differences were seen in the cost-shifting relationships before and after the BBA was passed and its effects felt in the industry. In addition, the intensity of hospital competition during this period did not significantly affect cost shifting, so that the degree of cost shifting observed did not vary between markets operating at different levels of competition.
All hospital types that exist nationally are present in California. Exhibit 1  compares characteristics of California hospitals with those of the nation. California hospitals, on average, are similar to all U.S. hospitals, except that higher proportions are for-profit and urban.
To measure a hospital’s financial condition, we calculated annual revenue and costs for each payer category, and we plotted the ratio of revenue (net of discounts) and costs over time (Exhibit 2 ). There is a striking and clearly visible inverse relationship between changes in this ratio for private-pay patients and changes in corresponding ratios for patients covered by public insurance. Until 1997, the general increase in payments relative to costs for Medicare and Medicaid coincided with a reduction of the margin for private payers. After 1997, the trend reversed. The Medicare margin (from 1997) and the Medicaid margin (from 1999) decreased; over the same time period, the private payer margin increased. To help verify this visual impression and to assess the overall strength of the aggregate relationship between private and public payment rates—controlling for costs—we regressed the logarithm of the average ratio of revenue to costs over the 1993–2001 period against the corresponding log-transformed Medicare and Medicaid ratios—a simple model based on nine data points, one dependent variable, and two independent variables. The estimated coefficients were large and highly significant, with values of approximately –0.68 for Medicare and –0.82 for Medicaid. Of course, such a limited aggregate analysis is subject to potential biases, which is why this paper’s key results are derived from a disaggregate hospital model.
We undertook the multivariate hospital-level fixed-effects analysis as a more stringent test of cost shifting during the period studied. Private payers’ price-elasticity coefficients obtained from the regression model were –0.17 for Medicare (p < .01) and –0.037 for Medicaid (p < .01). This means that controlling for average costs, for each 10 percent decrease in Medicare price (controlling for costs), there was a 1.7 percent increase in prices to private payers, and, similarly, for every 10 percent decrease in Medicaid price, there was a corresponding increase of 0.37 percent for private payers. The larger impact of changes in Medicare compared with Medicaid reimbursement could reflect the fact that for most hospitals, Medicare revenue is much larger than Medicaid revenue. These results suggest that one of the ways in which hospitals tend to respond to relative reductions in Medicare or Medicaid payments is by increasing private payers’ prices. We also found that even after many years, competitive pressures continued to reduce the growth rate of private insurers’ prices, independent of any cost shifting.
The relevant parameters were converted into dollar figures for the year 2001. As can be seen in Exhibit 2
Our analysis found a statistically significant and inverse relationship between changes in Medicare prices and private payers’ prices, controlling for costs, and a smaller inverse relationship between changes in Medicaid prices and those prices. Applying these relationships to the observed changes in average Medicare and Medicaid prices during the 1997–2001 period accounts for more than 12 percent of the total increase in private payers’ prices during that period. Several of our findings suggest that the standard economic theory of hospital behavior regarding cost shifting needs to be modified. We did not find that for-profit and not-for-profit hospitals had significantly different price relationships, even though we would have expected for-profit hospitals to have prices that are at, or close to, profit-maximizing levels prior to, and independent of, any changes in Medicare reimbursement.9 Furthermore, although private insurers appeared to have used hospital competition successfully to reduce the rate of growth in their payments, we found no statistically significant systematic relationship between the competitiveness of a hospital’s market and the strength of the relationship between prices paid by privately insured patients and those paid by Medicare (or Medicaid) patients. Both of these findings are consistent with those we found previously, suggesting that existing theories of hospital behavior are deficient. A major factor motivating this study was the expectation that hospital pricing behavior had changed with the implementation of the BBA. In fact, we found that the overall relationship between relative changes in Medicare and Medicaid prices and changes in private insurance prices was remarkably stable over both periods of relatively generous Medicare and Medicaid reimbursement before the BBA was enacted and of relatively constrained payment increases thereafter. Study limitations. Although we use our fixed-effects model for estimating the overall magnitude of cost shifting, for a variety of reasons these estimates might be biased downward to a significant degree. First, our definition of private patients includes Medicare and Medicaid patients covered through a managed care plan. This is advantageous in one respect, because our estimates of the effects of changes in Medicare and Medicaid prices are in no way threatened by potential biases. Since hospital payment levels negotiated by managed care plans for their Medicare patients are likely to be closely related to the prices negotiated for their privately insured patients, statistical biases could arise if payments for Medicare managed care were combined with those from Medicare fee-for-service patients. By excluding Medicare and Medicaid managed care patients, the remaining public payer revenues became exogenous. But this gain in analytical clarity came at the cost of a potential downward bias in the estimate of the relationship between private and public payers’ prices. The BBA reduced reimbursement rates to health maintenance organizations (HMOs) as well as to hospitals. These HMOs were then likely to respond by reducing their payments to hospitals for their Medicare patients. As a result, the increase in total payments from private insurers would appear lower. A second limitation is that the modeling approach might be generating estimates that are inherently understated. A fixed-effects model captures only the relationship between changes in a hospital’s public prices and changes in its private prices in the same year. As a result, this approach does not fully capture more complex adjustments to payment rates that take place over an extended period of time. For example, private payers’ prices cannot always be changed in the short run because of multiyear contracts; this delayed effect would not be captured with a fixed-effects model that relates yearly changes in the independent and dependent variables. Third, private price changes influenced by prices charged by other neighboring hospitals would not be fully captured by our disaggregate model. These limitations could partially explain the far larger effects found in the aggregate analysis. As Yehuda Grunfeld and Zvi Griliches argued many years ago, aggregate models might provide more accurate estimates when behavior is influenced by both time lags and other agents in the market, but one is not confident of one’s ability to properly specify all of these complexities in a disaggregate model.10 Finally, the BBA spelled out multiyear reductions in reimbursements; hospitals are likely to plan adjustments in their prices in anticipation of these changes in their payment rates.
A key issue is the generalizability of a study based on one state. As Exhibit 1 Policy implications. Two issues need to be clarified for policymakers. The first is causality. Thus far, we have been careful to assert no more than that we have observed a "relationship" between changes in Medicare and private payers’ prices. Is this "relationship" no more than an accidental correlation, or is it more likely that the changes in the Medicare and Medicaid payment levels caused the observed changes in private payers’ prices? Causality is notoriously difficult to prove in the social sciences, especially given the complexity of the environment in which hospitals operate and of their responses. At best, then, we must arrive at an assessment of the balance of probabilities. From this perspective, the results of this study suggest that one of the responses to changes in Medicare payment levels that hospitals tended to use were increases in the payment levels of their privately insured patients. This assessment is based on the fact that both the aggregate and hospital-level analyses agree that there was a strong inverse relationship between public and private reimbursement levels, controlling for costs. Finding such an inverse relationship is especially striking when using a hospital fixed-effects specification, since it isolates changes in the same year—that is, hospitals with a bigger decrease in average payments from Medicare in a given year (controlling for costs) tended to have a bigger increase in their private-pay prices. The stability of this relationship through many different specifications is highly suggestive of an enduring relationship. Since the changes in Medicare prices are almost entirely outside of a hospital’s control, it is hard to conceive of a mechanism, other than a causal relationship, that would explain this strong, stable, and statistically significant relationship. The second issue is the complex web of effects stemming from changes in reimbursement levels. Hospitals can be expected to respond to such changes along multiple dimensions. Reductions in Medicare reimbursement levels relative to costs could lead hospitals to reduce the quality of their care (through lower staffing ratios, for example), increase their efficiency, change their service mix, reduce uncompensated care, or increase prices to the privately insured; finally, they might have to accept lower profitability.11 Cost shifting is likely to be only one of their responses. This is clearly the case for California during the study period, since the increases in prices to the privately insured did not fully compensate for the relative reductions in payments from public programs. Another observed policy-relevant response to reductions in Medicare payment levels is reductions to the rate of growth in hospital costs.12 Federal and state policymakers must consider this web of interrelationships when making budgetary decisions. The results of this study add to this complexity. Policymakers, in considering the implications of decreasing Medicare and Medicaid payment rates to health care providers, must include the likelihood that some of this reduction will result in higher payment rates by private payers. In turn, higher hospital payments will tend to increase health insurance premiums and reduce private insurance coverage. Policymakers have the unenviable tasks of balancing these serious disadvantages with their goal of reducing the federal budget deficit.
Jack Zwanziger (jzwanzig{at}uic.edu) is professor and director of the Division of Health Policy and Administration and director of the Center for Health Services Research at the University of Illinois at Chicago School of Public Health. Anil Bamezai is a resident consultant at RAND in Santa Monica, California. This study was made possible by a grant from the Blue Cross Blue Shield Association (BCBSA). The BCBSA had no role in planning the investigation or in the data analyses and interpretation. The BCBSA was given a prepublication copy of the manuscript but had no approval rights. The paper was prepared with the assistance of BioMedCom Consultants Inc., in Montreal, Quebec.
This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Top
Data Sources And Methods
Study Results
