The Incidental Economist

In a comment yesterday, steve wrote,

Has anyone looked at the issue of market power in the context of rural vs urban providers? Most of the cases I have seen cited seem to identify large urban facilities, often university affiliated, as market powers. I would think that isolated, rural hospitals would have the same negotiating power, yet my experience would indicate that it seldom works that way.

It just so happens that today I was reading “Can Hospitals And Physicians Shift The Effects Of Cuts In Medicare Reimbursement To Private Payers?” by Paul Ginsberg in which he wrote,

The potential for cost shifting is also likely to vary geographically. Hospitals probably have more market power in smaller communities because concentration tends to be higher. Providers in small communities also might face more effective pressures to keep rates as low as possible. Both combined suggest that cost shifting has the potential to be more extensive. Indeed, many in contact with the insurance industry have noted sharp increases in rural hospital payment rates to private insurers in response to Medicare payment rate reductions resulting from the 1997 BBA. Rural hospitals, which tend to be natural monopolies, could have held rates way below their potential, so that when Medicare payment rates fell in relation to their costs, they could then raise rates to private insurers.

By the way, Ginsberg’s paper does a very nice job of laying out the most basic variants of cost shifting theory. There are more nuanced and complex theories out there that he doesn’t describe (not a critique, just a fact).

Jason Shafrin summarizes a 1999 paper on physician cost shifting from Medicare to private payers: Thomas Rice, Sally C. Stearns, Donald E. Pathman, Susan DesHarnais, Michelle Brasure and Ming Tai-Seale (1999) “A Tale of Two Bounties: The Impact of Competing Fees on Physician Behavior“, Journal of Health Politics, Policy and Law 24(6):1307-1330.

Shafrin writes,

The study examines the impact of [reduced Medicare physician payment rates] … using data from 1988-1991. …

[T]he authors find that Medicare fee reductions increased the volume of privately insured service only in some cases.  “Of the seventeen procedures groups, twelve had the expected negative signs and seven of the twelve were statistically significant at the 5 percent or 10 percent level…”  Additionally, it appears that in some cases, a decrease in Medicare fees increases private-pay services even for procedures unrelated to those that experienced a Medicare fee decrease.  For instance, when Medicare cut cataract reimbursement rates, ophthalmologists supplied more services to private-pay patients, but not necessarily more cataract surgeries.

Shafrin didn’t state what the level of cost shifting was in the cases in which it was found. Perhaps the paper doesn’t explicitly say so either. (I haven’t read it, though it is in my pile.)

In the health care cost shifting debate there are two hypotheses. One is that lower Medicare reimbursements motivate hospitals to seek higher payments from private payers. That’s the classic and pervasive notion of cost shifting. The other hypothesis is that hospitals with high degrees of power command high prices from private payers. This permits such well-paid hospitals to have weak cost controls, resulting in low or negative Medicare margins. That’s a somewhat counter-intuitive story that has been offered by MedPAC in reports to congress and explored in a new Health Affairs paper by three members of MedPAC’s staff, Jeffrey Stensland, Zachary Gaumer, and Mark Miller (summary on the Health Affairs blog).

Which hypothesis seems more likely to be correct? Do low Medicare prices cause high private payments, or do high private payments cause low Medicare margins (via relaxed cost controls)? There is no way to tell from descriptive analysis of observational data. The best evidence would come from a randomized trial. Go ahead and wait for it if you like, but it won’t happen. We can’t randomize hospitals to low and high payments any more than we can randomize them to low and high market clout.

The best we can do is look for natural experiments that can be exploited by well-designed observational studies. Sometimes there is exogenous (uncorrelated with private payment) variation in Medicare payment, such as that induced by the 1997 Balanced Budget Act. In a credible and well designed study, Vivian Wu exploited that phenomenon to deduce that, on average 21% of Medicare payment reductions are shifted to private payers. She also found that market concentration mattered, that in markets with the most dominant hospitals cost shifting rates were as low as 5%. (I reviewed Wu’s paper in a prior post.)

Wu’s results are consistent with other work, and I’m generally satisfied that cost shifting from public to private payers does occur, but at a level much lower than claimed by the hospital or insurance industries. However, that does not mean MedPAC’s hypothesis is incorrect. In fact Wu’s results actually strengthen it. In truth (or so I believe) payer-specific revenues, costs, and market power are, at least in part, simultaneously determined. There is no causal chain that runs only one way or another. Relatively dominant hospitals do cost shift (in the classic sense found by Wu, though at a relatively low rate) and they are also able to accommodate high cost structures and low/negative Medicare margins (as per the MedPAC interpretation). There’s really no disagreement between the two views.

Of course it would be very nice to see a convincing study that explores the issue explicitly from the MedPAC perspective. As the authors make clear themselves, the paper by Stensland, Gaumer, and Miller only illuminates the hypothesis and shows that it might plausibly be true. But it does not show evidence that is necessarily consistent with a causal connection between market power and costs. That ‘s not a critique, just a fact.

Theory-minded economists might dismiss the notion that an organization would allow revenue to drive costs. Don’t all organizations minimize costs to maximize profit, independent of revenue? Stensland, Gaumer, and Miller think that nonprofit hospitals would not.

When nonprofit hospitals have more resources, they tend to spend those resources because nonprofit hospitals do not have shareholders to distribute profits to. The nonprofit hospital’s expenditures could be on service-line expansions, such as a new cardiac surgery wing; on acquiring physician practices; on patient amenities, such as larger rooms; or on other capital expenditures that help the hospital maintain and expand its market share of private-payer patients.

On the other hand, in theory for-profit hospitals should minimize costs irrespective of revenue and should maximize revenue over each payer independently. For such hospitals, neither cost shifting theory should hold. If costs don’t vary with revenue then they can’t explain Medicare margins. And a revenue maximizing firm cannot compensate for low Medicare payment with high private payment because, as for profit entities, they’re already maximizing private payment independent of other revenue sources, including Medicare.

Well, that’s theory. The world is often messier. In her empirical study Wu did not find a statistically significant relationship between cost shifting and hospital profit status writing that, “cost shifting is not determined solely by institutional characteristics.”

Market power, costs, private, and public payment are almost surely all related and nothing definitive can be learned without exogenous variation in at least one of these factors. Though work that exploits just that exists, the cost shifting debate will no doubt continue, due in part to the allure of descriptive work based on the weak assumption that costs are exogenous (i.e. outside the control of administrators). But one thing ought to be settled. When two things are simultaneously determined it cannot be said which causes which. Do low Medicare margins cause higher private payments or vice versa? The answer is yes (but to a small degree). And the key mediating factor is market power.

Administrative vs. competitive mechanisms constitute a fundamental dichotomy in health care pricing. For the former think Medicare, for which prices are largely set via political/administrative processes. For the latter think the private non-elderly market, for which prices are set via insurer-provider negotiation. In a recent paper in the American Journal of Managed Care, Chernew, Sabik, Chandra, Gibson, and Newhouse shed some light on the implications for health care prices of administrative vs. competitive pricing. The results are not surprising.

In a retrospective descriptive analysis of geographic variation the authors find that Medicare and large-firm commercial hospital utilization were positively correlated, but spending was not. The authors interpret these results with appropriate caution since they are correlations. However, they are consistent with other work that suggests Medicare and commercial insurers have different responses to hospital competition. Commercial insurers can exploit it to drive costs downward. In contrast, Medicare may be less influenced by the effects of provider competition since it doesn’t negotiate prices (or doesn’t do so in the same fashion as commercial insurers).

The authors conclude,

The potential susceptibility of private payers to provider market power has important implications when assessing the merits of private markets or public markets in setting prices. Administrative price systems have many flaws, which are fundamentally related to the difficulty in determining the appropriate price when costs are heterogeneous, are not known very precisely, are changing over time, and may reflect discretionary provider behavior.

… Yet despite all the concerns about administrative pricing, our analysis appears to suggest that administratively set prices seem to reduce purchaser vulnerability to provider market power. The challenge for policymakers interested in administered prices must be how to mitigate distortions in the price-setting process, although policymakers will never have enough information to establish perfect (economically efficient) prices (bundled or otherwise).

The analogous challenge for policymakers interested in market systems is how to avoid the pitfalls associated with provider market power. It is not clear whether concerns about market systems are more important or will be easier to mitigate than concerns about administered pricing. However, as the country moves forward with changing the healthcare system, these concerns will be paramount.

That’s an even-handed take on the two pricing systems. Neither is perfect from every perspective. Proponents of one can (and do) easily point to flaws in the other. However, given that our system is and will remain a mix of public and private payers, sound policy must attempt to address the issues raised by both.

Before concluding, I also want to highlight what Chernew, et al. say about cost shifting:

[O]ur analysis does not necessarily indicate cost shifting. The pattern of results we observed, particularly the association with market structure, may merely reflect differential market power as opposed to a causal relationship between prices in different sectors.

That market structure has an important impact on the degree of cost shifting has already been covered on this blog. Chernew, et al.’s paper is just one more in a large body of work that suggests that ignoring market structure when considering health care pricing and price dynamics misses the point.

In her careful study of the effects of Medicare cuts in the late 1990s on private prices for hospital services Vivian Wu obtained an estimate of a 21% cost shift. That is, when Medicare prices went down by $1, private prices went up in response by $0.21. She also found the degree of cost shift varied by level of hospital competition. For the most competitive markets, cost shifting could be as low as 5%.

What are we to make of this 21% figure? Is it the maximum degree of cost shifting we might expect due to Medicare cuts? Or is it the minimum? Where in the range of likely levels of cost shifting does this point estimate sit? Without further information reasonable people can disagree. In her reaction to Wu’s study Megan McArdle took a stab at this question, speculating that the 21% was not a ceiling. She declared it a floor.

McArdle is “wont to prefer academic studies,” as am I. So let’s look at some. I am aware of several that estimate levels of cost shifting below Wu’s 21% figure. For instance, Cutler found no evidence of cost shifting in the 1990s (0%) and Zwanziger and Bamezai estimated a cost shift rate of 17%. To be sure, one can find studies that obtain estimates higher than Wu’s 21% figure, though many are based on even older data and some use indefensible methodology.

So, I agree with McArdle that we cannot conclude with full confidence that 21% is a ceiling. However, I disagree that we can confidently conclude it is a floor. After reviewing the literature I think a reasonable person can only conclude that it is a credible point estimate far closer to the truth than those suggested by the industry.

Good, policy relevant, health economics studies are hard to do. They have to satisfy many criteria in multiple dimensions, being fundable, tractable, publishable, timely, relevant, built on accepted methods appropriate for the application, and for which data are available. Moreover, given the delays in funding, the time it takes to do the work, and the slow journal review process one must anticipate by years what is likely to be important. So, it is hard. But when it is done well it is a thing of (wonky) beauty.

Lately there have been quite a few health care cost shifting studies that fail to be beautiful, even in the eye of a wonk. In health care “cost shifting” is the idea that lower public payments to providers lead (causally) to higher private payments and health insurance premiums. Not too long ago I described a recent cost shift study and traced its assumptions back to their source. I found that it was built on a weak foundation and therefore didn’t advance our understanding of cost shifting. From poor assumptions and inappropriate methods incorrect results follow, leading to the wrong policy recommendations.

This is, of course, relevant. The notion that most or all of public payment reductions are or will be shifted to private payers is pervasive. With Medicare expansion part of the Senate public option compromise there are renewed claims from hospital and physician groups that low payment from Medicare leads to higher private payments and insurance rates.

Fortunately, a sound cost shifting study by Vivian Wu has recently been published (Hospital Cost Shifting Revisited: New Evidence from the Balanced Budget Act of 1997, International Journal of Health Care Finance and Economics, 12 August 2009). It satisfies the criteria of a good study and leads to sensible policy recommendations consistent with those implied by other studies. Wu’s is just the latest in a growing body of work suggesting that increasing competition among hospitals should be part of an overall strategy to control health care costs.

Wu’s main result is that on average prices paid to hospitals by private payers increases by 21 cents in response to each dollar reduction in public revenue. By way of comparison, this 21% rate of cost shift is about half of the lowest estimates produced by industry studies and is far below their common assumptions of 50% to 100%.

If you can stand a bit of economic, econometric, and public policy wonkery read on. The details reveal just how good a job Wu did in her study and why others are not up to the highest standards of good technique and scholarship. First, Wu describes the two schools of thought on cost shifting. One espouses the “market power” hypothesis which is that hospitals with large shares of private patients are able to cost shift more due to their greater bargaining power. The other abides by the “strategy” hypothesis leading to the notion that hospitals with a greater share of private payers cost shift less because they are relatively more immune to changes in public payments. (Here “strategy” refers to that of the public programs, which includes free riding on high private payments.)

Next Wu thoroughly reviews the academic cost shifting literature, some of which I’ve covered in prior posts (under the cost shift tag). In doing so she makes the crucial point that changes in private and public revenues are endogenous, as is revealed directly by the “strategy” school argument, though there are other reasons. Quoting Wu’s paper,

On the one hand, Medicare prices can affect private prices because the latter are negotiated after observing Medicare prices. On the other hand, Medicare revenues can be affected by private prices as well. Medicare, knowing private payers’ generosity, can strategically set prices too low to free ride.

Hence, one cannot conclude anything about cost shifting by direct examination of private and Medicare revenues, a mistake made in many studies. One must use more sophisticated techniques (instrumental variables to use precise jargon) to relate changes in private revenues to the exogenous component of public revenue changes.

Of course, having made this point Wu applies appropriate methodology to estimate a (plausibly) causal relationship between private price changes and changes in hospital Medicare revenue due to the 1997 Balanced Budget Act (over the period 1996-2000). In addition to finding a 21% cost shift she also finds evidence that the “market power” hypothesis dominates. That is, if the “strategy” hypothesis holds it is weaker and therefore overwhelmed by market power forces. To put this more simply, in a more competitive hospital market one should expect less cost shifting.

The policy implications are clear. Wu doesn’t state them, but I will. Within the range of variation studied by Wu, with respect to hospital payments, overall health costs can be reduced by 79 cents per dollar of Medicare payment reduction, the other 21 cents being shifted to the private sector. However, the more competitive the hospital market the less the cost shift. For some hospitals in some markets Wu found cost shifting rates as low as 5%. Therefore, sound public policy would encourage greater competition among providers (wherever possible) in tandem with reductions in public payments. Doing both concurrently would reduce public health expenditures with minimal impact on private payments.

Recently I wrote about the September 2009 Health Affairs paper by Dobson et al. “How A New ‘Public Plan’ Could Affect Hospitals’ Finances And Private Insurance Premiums.” In that paper the authors assumed that hospitals would shift 50% of reductions in payment from public payers to private ones. Their conclusion that private premiums would skyrocket follows necessarily from this assumption. It turns out the 50% assumption rests on 14 year-old analysis of 18 year-old data. In this post I follow the paper trail back to the source document.

As support for the 50% cost shifting assumption, Dobson et al. cite the March 2009 Lewin report by Sheils titled The Health Benefits Simulation Model (HBSM): Methodology and Assumptions. End note 7 of Dobson et al. reads, “Sheils estimates the cost shift at 40 percent, which is roughly consistent with our assumption.” By “roughly” the authors mean that 50% is roughly 40% (a mere 25% difference).

In the HBSM report Sheils justifies the 40% cost shift assumption as follows, “Our own analysis of hospital data indicates that about 40 percent of the increase in hospital payment shortfalls (i.e., revenues minus costs) in public programs were passed-on to private-payers in the form of the cost shift during the years studied.” A footnote indicates that their “own analysis” is that provided in a December 1995 report and “years studied” means 1991-1992. The 1995 report referenced is one by Sheils and Claxton titled “Potential Cost Shifting Under Proposed Funding Reductions for Medicare and Medicaid: The Budget Reconciliation Act of 1995.”

If the 1995 Sheils and Claxton paper is online I couldn’t find it. But I am a curious fellow with urgent questions. What analysis in 1995 revealed a 40% cost shift? Why is the document so hard to find, yet so frequently cited (18 hits in a Google search of the title, many to documents written or testimony given within the past few years)? Could I get my hands on this old study? And what would I find when I pulled back the curtain?

While neither of the authors could provide the report (Claxton, now Vice President of The Henry J. Kaiser Family Foundation returned my e-mails, Sheils did not), I obtained a PDF on request by calling The Lewin Group directly (703-269-5500). Sure enough, it included a calculation that had a 40% bottom line. Better yet, it included enough aggregate hospital cost and revenue data (from the American Hospital Association’s survey of hospitals) to replicate the calculation. (See the “wonk note” below for a summary of the calculation.)

There is a key assumption embedded in the calculation, that costs are beyond control of hospital managers. Working through the calculation, every additional dollar of cost produces an additional dollar of public to private cost shift. Is it credible that costs drive private revenue and not the other way around (at least in part)?

Another interesting fact revealed by the data is that that profits increased as the purported cost shifting was taking place. In fact, profit increases represent two-thirds of the increase in payments from private payers. One way to look at it is that when the hospitals were done being forced into increasing charges to private payers due to public shortfalls they were then forced to increase them a lot more to support growing profit.

If cost shifting exists the approach in the 1995 Lewin report does not make a strong case for it. An assumption of inviolable costs is not reasonable.  Profits increased at the expense of private payers and despite lower public payments, suggesting insufficient pressure to reduce costs. One thing is clear. The document tells us a little bit about how to use early 1990s hospital cost report figures to make it appear as if a 40% cost shift is an inevitable, causal phenomenon. But that’s a poor foundation for the analyses of health reform conducted in 2009 that build upon the 1995 results. Like the Wizard of Oz, what was behind the curtain was rather old and less powerful than suggested.

---

Wonk note: The methodology of the 1995 Lewin report that derives a 40% cost shift is as follows. The difference between the 1992 and 1991 non-private (I’ll call it “public”) payer shortfall (revenue less costs) is computed (call this value A).

Next, the 1992 private payer revenue less its expected value is calculated (call this value B). The expected value of the 1992 private payer revenue is the product of the 1991 revenue-to-cost ratio and the 1992 cost.

From B a change from expected 1992 profit is subtracted (call the difference C). Expected 1992 profit is the product of the 1991 profit-to-revenue ratio times the 1992 revenue. The cost shift is the ratio C/A which works out to 40% using the actual figures in the report.

Recently renewed claims by the insurance industry (or consulting firms on their behalf) that lower public payments to health providers result in higher health insurance premiums have reinvigorated the debate over cost shifting. This is no mere academic argument. Whether or not cost controls implemented in public health programs actually reduce overall health expenditures is central to how we finance health care and the range of options available to control its costs.

So it is important that we get straight in our minds the extent to which cost shifting exists, if at all. In an earlier post I reviewed academic literature on cost shifting and concluded that it either doesn’t exist or if it does it is small. Studies have found evidence of a 0.4 to 1.7 percent increase in private payments in response to a 10 percent decrease in Medicare and Medicaid fees. Older studies found higher rates but under a 1980s health care system that differs considerably from what exists today.

Thus, the most recent academic literature suggests at most a degree of cost shifting well below the level others have claimed. But, to be thorough, a few weeks ago I issued a challenge to readers: send me any credible peer-reviewed article that provides evidence of a substantial level of cost shifting.

It turns out the journal Health Affairs recently published an article pertaining to cost shifting. “How A New ‘Public Plan’ Could Affect Hospitals’ Finances And Private Insurance Premiums” by Allen Dobson, Joan DaVanzo, Audrey El-Gamil, and Gregory Berger (all employees of Dobson DaVanzo & Associates, LLC, a health care consulting firm) appeared on September 15, 2009. Some of the same authors published an article in Health Affairs in 2006 with much of the same intellectual content (at the time they were employed by the Lewin Group).

In a press release the authors summarize their findings that “a government-run plan that is aggressively implemented to include large proportions of the privately insured could test the U.S. health care financing system…Rising hospital private-payer payment-to-cost ratios could be followed by rising private insurance premiums.” In the paper itself, the key driver of this finding is the assumption that cost shifting will occur. The authors write, “Our model assumed that hospitals shift costs to private payers at a 50 percent rate.” As support, they cite Lewin and Milliman reports, the latter by the same individuals who produced the one I previously reviewed.

That is, the authors’ conclusion is preordained by their assumption. Costs will be shifted at a 50% rate because it is assumed they will be. No serious economist I know of believes this assumption. And nobody should if it is not based on a body of credible, peer-reviewed work, which it is not.

Why does the cost shifting fallacy persist? First, it stems from an accidental or intentional confusion with the notion with price differentials. That payer A pays less than payer B (a price differential) does not imply that payer A’s lower payment caused payer B’s higher one (cost shifting). Second, it is (or has been) an argument of convenience for the insurance and hospital industries. A legislator believing the argument would be less inclined to exert downward pressure on the public health budget for fear of constituency backlash in reaction to higher private payments and premiums.

Today, it is hard to understand how the cost shift argument, even if true, could work in industry’s favor. As Uwe Reinhardt put it in his NY Times Economix piece on this topic, “if [hospitals and health insurers] cannot resist a cost shift from government insurance programs to the private sector caused by a cut in, say, fees, then presumably they cannot resist a reduction in the growth of government spending on health care for any reason.” Why then should we build a health care system around an industry that claims it cannot solve the very problem health reform is, in the long run, intended to address?

It is time to recognize the 50%-100% health care cost shifting assumption for what it is and send it the way of the fallacies of antiquity. Like the flat Earth or the Ptolemaic model of the universe, it belongs in the dustbin, at least until such time as its existence is demonstrated with by a body of work, conducted by sound methods, and passing peer-review. Today, as far as I know, no such credible body of work exists, and the new Health Affairs paper by Dobson et al. makes no contribution toward one.

Here’s an instance where I don’t really need to do any blogging. Others I respect and agree with have already thoroughly read and commented on the new PricewaterhouseCoopers (PWC) report that AHIP is using as a basis for argument against the Senate’s health reform proposal. So, if you care to know what I think about it, please see Aaron Carroll’s analysis or that of Ezra Klein. (And subscribe to both. They’re consistently good.)

And, I’ve already written a great deal about cost shifting, which is relevant to the PWC report.

In response to my two posts on cost shifting (or lack thereof) in health care, I’ve heard many arguments why I am wrong (see comments at The Health Care Blog and Ezra Klein’s blog). In brief: my claim is that there is little or no cost shifting from public to private payers in health care. I cited recent studies (which in turn cited peer-reviewed papers) to support my claim.

But nobody has yet pointed to credible studies that counter what I wrote. Here’s the challenge: find a qualified (see below) publication that supports your health care cost shifting claim. Send it to me (pdf or URL to paper please). I will read it, and I will blog about it.

“Qualified” means:

• Appears in a peer-reviewed journal (send me a link to the journal’s editorial policy that confirms this if it is not obvious) or is based in large part on peer-reviewed articles.
• Has been published in the last decade.
• Is by authors that are not obviously conflicted (not employed, recently employed, or funded by the insurance or hospital industries).
• Has not already been covered in my posts or the references I cite (see below).

References already covered (don’t bother to send me these):

• MedPAC report, March 2009.
• CBO report, December 2009.
• GAO report, August 2005.
• Milliman report, December 2008 (doesn’t actually qualify since there are conflicts of interest, but I reviewed it anyway because it is what the insurance and hospital industries cite).
• Richard Frank, Health Affairs, March/April 2001.
• Dranove, D., and W. White. 1998. Medicaid-dependent hospitals and their patients: How have they fared? Health Services Research 33, no. 2: 165–185. “We find no evidence that [CA] Medicaid-dependent hospitals raised prices to private patients in response to Medicaid (or Medicare) cutbacks; if anything, they lowered them…. We find no evidence of cost shifting.”
• Zwanziger, J., and A. Bamezai. 2006. Evidence of cost shifting in California hospitals. Health Affairs 25, no. 1: 197–203. “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.”
• Zwanziger, Glenn A. Melnick, and Anil Bamezai, Can Cost Shifting Continue in a Price Competitive Environment Health Economics, vol. 9, no. 3 (April 2000).
• Michael A. Morrisey, Cost Shifting in Health Care: Separating Evidence from Rhetoric (Washington, D.C.: AEI Press, 1994)
• Jack Hadley, Stephen Zuckerman, and Lisa I. Iezzoni, “Financial Pressure and Competition: Changes in Hospital Efficiency and Cost-Shifting Behavior,” Medical Care, vol. 34, no. 3 (1996), pp. 205–219.