MEDICARE

Joseph P. Newhouse Harvard University and NBER

Prepared for the Conference on Economic Policy During the 1990s Kennedy School of Government

June 27-30, 2001

I wish to thank David Cutler, Victor Fuchs, Michael O’Grady, Jon Gruber, and Bruce Vladeck for comments. Any remaining errors are my responsibility.

2

Medicare, which accounts for about an eighth of the federal budget, covers health

care costs for three groups of beneficiaries.1 The great bulk of the spending, approximately 85 percent, provides benefits for those over 65.2 About 10 percent covers those eligible for Disability Insurance, and the remaining five percent covers those of any age with end stage renal disease (kidney failure).3 Enacted in 1965 as part of the Great Society, the program was implemented in July 1966 and extended to the disabled and those with renal disease in 1972. It now enrolls approximately 40 million individuals.

Given that Medicare pays more than $5,000 per beneficiary, it should not be

surprising that it is a voting issue for the elderly and near elderly and hence an issue that no President or member of Congress can ignore. Of those voters 60 years of age and over who responded in exit polls following the 1996 presidential election, Medicare/Social Security ranked as the top issue; 29 percent ranked it first versus only 8 percent among voters under 30 (Blendon, et al., 1997). Of the 14 percent of voters who said Medicare/Social Security was the most important issue in deciding their vote, half were 60 and over and another quarter were aged 45 to 59 years. And the elderly vote disproportionately; 61 percent of those 65 and over voted in the 1994 Congressional elections versus 20 percent of those 18 to 24 and 32 percent of those 25 to 34 (Blendon, et al., 1995). Furthermore, three in every four adults of all ages respond in polls that Medicare is important to their families (Blendon, et al., 1995).

In this chapter I first describe Medicare as the 1990s began and then discuss some

of the important changes made to the program in the 1990s. I next analyze current issues, dividing them into relatively short-term issues of reimbursement methods, longer-run issues of financing, and potential future benefit expansions. I conclude by commenting on certain administrative issues. A more extensive economic analysis of many of the issues discussed below can be found in (Newhouse, 2002). Because Medicare is complicated, describing it and the changes to it during the 1990s does not always make for easy reading. Nonetheless, I have tried to minimize the detail.

Much of the public debate of the last several years has been about how to finance

Medicare a decade or more hence when the baby boomers start to swell the ranks of the beneficiaries. Although I touch on this issue, I focus more on Medicare in the here and now. Medicare relies on administered price systems that are not likely to see us through the next decade. But improving those systems raise difficult substantive and political issues.

1 The one-eighth figure includes the portion of Part B payments financed by beneficiary premiums. If they are considered outside the government budget, Medicare accounts for about 11 percent of outlays. 2 Medicare is a secondary payer for those over 65 actively employed who have access to group insurance through their employer. Those few elderly who are not eligible for Social Security must pay a premium for Medicare Part A. 3 Medicare eligibility for the disabled begins two years after eligibility for disability insurance. Those under 65 with renal disease, if covered by employment-based insurance, are to be covered through their employer-based coverage for the first 30 months of eligibility.

3

My vantage point on Medicare comes from serving on the various Commissions the Congress created to advise it with respect to Medicare. I was a Commissioner of the Physician Payment Review Commission from 1993 to 1996 and the chair of the Prospective Payment Assessment Commission in 1996 and 1997. The Balanced Budget Act (BBA) of 1997 combined these two Commissions into the Medicare Payment Advisory Commission (MedPAC), and I have served on that Commission from 1997 to the present. As a result of this experience, my viewpoint likely gives a larger role to the Congress and a lesser role to the Executive Branch than might a chapter written by someone who served in the Executive Branch. Nonetheless, the Congress clearly writes rather detailed policy into statute and is thus an important actor vis-à-vis Medicare.

Medicare at the Beginning of the 1990s Benefits. The Medicare insurance contract remains patterned after the indemnity

policies that prevailed in the 1960s. Part A, an entitlement financed from payroll taxes, covers services of institutional providers, most notably hospitals and skilled nursing facilities.4 It accounts for roughly 60 percent of Medicare spending. Part B is a voluntary insurance program covering physician and other outpatient services. It is financed 75 percent from general revenues and 25 percent from premiums paid by the elderly; given this degree of subsidy, over 95 percent of the elderly purchase Part B.5 Importantly, Medicare was designed to cover the cost of acute medical services; it was not intended to cover chronic long-term care. As we shall see, however, in practice this distinction has become somewhat blurred.

Reflecting the insurance policies of the 1960s, Medicare does not cover outpatient

prescription drugs, nor does it have a stop-loss feature that limits a beneficiary’s out-of-pocket spending in a year.6 Hospital coverage, in fact, gives out entirely after a 90-day hospital stay.7 Hospital services have a deductible equal to the average cost of a day in the hospital ($792 in 2001), well above the deductible in almost all employer-provided policies for the under 65; Part B services have a $100 annual deductible and a 20 percent coinsurance rate. Home health services are an exception and have no cost sharing.

4 The payroll tax rate is currently 2.9 percent on all earnings, with half nominally paid by the employer. Prior to 1991 the upper limit on taxable earnings was the same as Social Security. OBRA90 raised the limit to $125,000 from $51,300, effective in 1991, and indexed the limit. The limit was removed entirely in OBRA93, effective in 1994. 5 The elderly with incomes below the federal poverty line have their Part B premiums as well as their cost sharing paid for by Medicaid, and there is also premium assistance (but no cost sharing assistance) for those between 100 and 120 percent of the poverty line. 6 Why private health insurance developed in this fashion historically is an interesting puzzle for economic theory. My guess is that actuaries initially feared to rate a policy with a stop-loss feature, and once policies without such a feature became established, firms feared adverse selection from introducing them. Nonetheless, most policies do now have a stop-feature or do not require one because cost sharing is so low. 7 Medicare covers 90 days of a hospital stay within an episode of illness. A new episode of illness begins after a beneficiary has been out of the hospital or the skilled nursing facility for 60 days. Beneficiaries have an additional 60 days of coverage from a one-time “lifetime” reserve.

4

Probably because Medicare lacks a stop-loss feature, a supplementary insurance industry has grown up. Over 90 percent of beneficiaries have some form of supplementary insurance that covers much of the cost sharing, thereby converting cost sharing at the point of service to premium payments. This supplementary insurance comes from four sources: employer-provided retiree health insurance; individually purchased coverage (“Medigap”); additional benefits provided to those who join a Health Maintenance Organization (HMO); and Medicaid (Table 1).8

Overall Medicare covers about two-thirds of expenses among those elderly living

in the community.9 Supplementary insurance, both employer-provided and individually-purchased, covers 11.5 percent, Medicaid covers 2.5 percent, and 15.2 percent is paid for out-of-pocket (Medicare Payment Advisory Commission, 1999b).10

Reimbursement. Medicare also emulated the insurance policies of the 1960s in

giving its beneficiaries freedom to choose among almost all providers, with little or no difference in price to the beneficiary. Moreover, Medicare’s designers did not want providers to bill beneficiaries additional amounts, or at least wanted to limit such “balance billing.”11 Not only could balance billing undermine freedom of choice, but it would defeat the entire purpose of Medicare if providers charged patients what they would have charged without Medicare and collected Medicare reimbursement in addition. If Medicare was to shoulder the bulk of the reimbursement load and there was to be freedom of choice, Medicare had to pay physicians and other providers at rates that virtually all of them would accept. Of course, Medicare could not agree to reimburse any price a provider named, and it therefore developed administered pricing systems that have grown steadily more elaborate. The word rococo comes to mind. Initially the program followed the methods of Blue Cross and Blue Shield insurance plans by reimbursing institutional providers such as hospitals its share of costs and physicians their usual fees, subject to an area wide ceiling.12

In FY1984 a major change occurred in hospital reimbursement, which increased

the incentive for efficient production (Wise and Woodbury, 1994), (Shleifer, 1985). Over a five-year transition the program changed to a Prospective Payment System (PPS), that is, changed from paying its share of hospital costs to a fixed or “prospective” amount per admission. The amount paid varied according to the patient’s diagnosis and whether certain procedures were performed; this information classified the patient into one of about 500 Diagnosis Related Groups (DRGs), to which a relative weight was attached. The Congress legislated a “conversion factor,” which translated the relative weight into a dollar figure, and which it “updated” or

8 Medicaid beneficiaries, of course, make no premium payments. 9 This includes the managed care enrollees. 10 The remaining 3.7 percent comes from other sources such as the Veterans Administration. 11 From the outset Medicare prohibited hospitals and other institutional providers from balance billing. In the 1980s physicians were prohibited from charging more then 10 percent above the fee schedule. 12 Medicare’s share of costs was defined as its share of patient days; costs were determined from audited cost reports filed by providers.

5

increased each year.13 The payment for each admission also varied with the hospital’s area wage index, the number of interns and residents per bed at the hospital, and to a small degree with the size of the city in which the hospital was located.14

Other institutional providers, including hospital outpatient departments, continued

to be reimbursed on the basis of cost, which not only offered no incentive for efficient production, but gave hospitals an incentive to adopt accounting conventions that shifted as much joint cost as possible from prospectively reimbursed inpatient services to other parts of the hospital, such as the outpatient department, the Skilled Nursing Facility (SNF), and the rehabilitation unit.15

HMO Reimbursement. When it began in 1966 Medicare had no mechanism for

paying HMOs a set amount per member per month, the method by which they were paid for their under 65 members. Rather, HMOs billed Medicare like any other provider, that is, on a fee-for-service basis.16 Not surprisingly HMOs did not like having to engage in billing efforts for Medicare that they did not use in their private business, but there was Congressional opposition, especially in the Senate, to contracting with HMOs at a fixed sum per member per month.17 Medicare enrollment in HMOs was modest; indeed, in 1966 when Medicare began even the enrollment of the under 65 in HMOs was modest.18

By the early 1980s the notion grew that HMOs might be an efficient means to

deliver care and that Medicare should be more accommodating toward them. The 1982 Tax Equity and Fiscal Responsibility Act (TEFRA) authorized the creation of a risk contract, under which HMOs would be reimbursed a fixed amount per member per month. This fixed amount, called the Adjusted Average per Capita Cost (AAPCC), was, for an average enrollee, 95 percent of traditional Medicare’s average payment in the enrollee’s county of residence. The five percent off-the-top reduction was taken so that Medicare could share in the assumed efficiencies of HMOs.

13 The conversion factor was the dollar amount paid for a DRG with a relative weight of 1.0. 14 For further discussion, see (McClellan, 1997). 15 In the case of the institutional providers not covered by the PPS I am greatly oversimplifying a set of complex reimbursement rules; Medicare sometimes reimbursed a function of a prior year’s cost per case updated for inflation (e.g., rehabilitation units), sometimes simply on the basis of cost (e.g., ancillary services in SNFs), and sometimes costs subject to maximums (e.g., routine costs in SNFs and home health agencies). 16 In 1972 an option was added for Part B services to be reimbursed on the basis of cost, and in 1982 this was extended to all services. These contracts are now being phased out. 17 Essentially all HMOs had the capability of billing fee-for-service for a few payers as Workers Compensation or if emergency services were rendered to a non-enrollee, but such billing was typically treated exceptionally. In 1976 Medicare began a demonstration project with risk contracting in one HMO ((Eggers, 1980) and (Eggers and Prihoda, 1982)). 18 I have not found data on the number of beneficiaries in HMOs in the early years, probably because HMOs were not distinguished from other providers in their billing. In fact, the term HMO was not even coined until 1971, although entities such as the Kaiser Health Plan had existed for several decades. But in the 1980s, the HMO Medicare market share was only 3 percent. It almost certainly was less than that at the outset of the program.

6

The AAPCC payment was adjusted for the enrollee’s age, sex, institutional status

(e.g., was the person’s residence a nursing home), and whether the person was eligible for Medicaid. For example, if 65-69 year old females, living at home and not on Medicaid, spent 93 percent as much as the average Medicare beneficiary, the HMO received 93 percent of the AAPCC amount for enrolling a person with those characteristics.19 In addition to the AAPCC, HMOs could charge their members a premium of up to the actuarial value of the cost sharing provisions in traditional Medicare. If HMOs could provide services for less than the AAPCC payment plus the allowable premium, they could reduce the premium or provide additional services to beneficiaries or both.20

Changes to Medicare in the 1990s The First Six Years Compared to the second half of the 1990s, the first part of the decade was

relatively quiet for Medicare. The major change was to how Medicare paid physicians.

Changes in Physician Payment. The OBRA 1989 legislation enacted a major

reform of physician payment, which began to be implemented in 1990. Physician spending had grown very rapidly since the inception of the program; between 1975 and 1990 it increased at an annual real rate of 9.7 percent. This growth largely stemmed from an increased quantity of services (Physician Payment Review Commission, 1990). By 1989 Part B spending, 75 percent of which was for physician services, had become the largest domestic program funded from general revenues. At a time of large deficits, the Congress sought additional constraints on Medicare physician spending. In addition, many felt that Medicare fees for procedures were too high relative to those for evaluation and management services such as taking a history. As a result, the Congress legislated a series of ad hoc reductions in certain “overpriced” procedures in the late 1980s.21

19 To obtain stable estimates of these adjustments, they were calculated at the national level. For the same reason the county average payment was computed as an average over a five-year period. 20 The applicable regulations invoked a concept of the Adjusted Community Rate (ACR), which was the rate charged by the HMO in their private business, adjusted for benefit and demographic differences with the Medicare population. If the cost of services to the Medicare population was less than the ACR, including a profit rate on the private business, the HMO was to provide additional benefits or refund the excess to the government. Needless to say, the government received few refunds. The adjustments in the ACR calculation have a considerable degree of arbitrariness and are probably not binding; even so, competition in metropolitan areas among HMOs forced any rents in HMO reimbursement to be passed through to beneficiaries in the form of additional benefits or lower premiums, as I come to below. 21 That Medicare was paying too much for some procedures was plausible. In the case of several newer high tech procedures, productivity had improved considerably, but Medicare’s administered prices were rigid downward and had no ready way of adjusting for these improvements. Among the overpriced procedures were coronary angiography and cataract operations.

7

The essence of the 1989 reform addressed both the issues of spending growth and the structure of relative prices across physician services. It put in place a formula, the Volume Performance System (VPS) that set a target for the total amount of money Medicare would pay to physicians in a year. Under the VPS a target increase (“performance standard”) in physician spending was set. The target was a function of a five-year moving average of the annual increase in the quantity of physician services, on the grounds that this value would reflect scientific and technical advances for which Medicare should pay. If the quantity of services (“volume”) increased above its five-year trend, unit prices two years later (the “conversion factor”) would be proportionately decreased to constrain total spending on physician services. Conversely, a fall in volume below the five-year trend caused unit prices to rise in the short run. As volume stayed low, however, the five-year moving average would start to fall, and the target would fall, thereby lowering future updates. Although it was not realized at the time, this method was to lead to substantial instability in physician fees.

In effect, the Congress had budgeted an amount to be spent on physician services,

whereas before the VPS spending was simply the product of whatever services physicians ordered times Medicare fees. As can be seen in Figure 1, the VPS was effective in constraining spending growth on physician services. The real growth rate in spending declined from the 9.7 percent rate of the prior 15 years to 2.8 percent between 1990 and 1997.

In addition to the VPS, Medicare adopted a new schedule of relative prices for

physician services that reflected the amount of “work” for each service ((Hsiao, et al., 1988),(Hsiao, et al., 1992a),(Hsiao, et al., 1992b)). The new relative fees resulted from consensus among physicians within a specialty on relative fees within that specialty. Relative fees across specialties were aligned through so-called linking procedures, which were either the identical procedure performed by two different specialties (e.g., laminectomy performed by orthopedic and neurosurgeons) or different procedures that a panel of physicians regarded as equivalent. Prices for linking procedures were to be as equal as possible, thereby aligning the fee scales for the different specialties.22 The resulting relative prices implied a substantial redistribution across specialties. In particular, procedure oriented specialties, which included surgeons as well as certain medical subspecialists such as invasive cardiologists, were to have fees for many of the services they performed reduced, whereas primary care physicians were to receive higher fees.

Not surprisingly the losers resisted the reform. As a result, the potential

redistribution was initially mitigated in two ways. First, the implementation of the new fee schedule (but not the VPS) was put off for three years, until 1992, after which time there was a four-year (linear) transition to the new fee schedule, so that the new relative prices were not fully in place until 1996. Second, “practice costs” were excluded from the reform, and continued to be passed through under the earlier

22 With one linking procedure, the different specialty scales could be aligned exactly; with multiple linking procedures, scales were aligned to minimize the sum of squared errors in the prices.

8

reimbursement rules. Practice costs refer to the approximately half of physician revenue that is not net physician income; that is, they encompass “overhead” costs such as office rent, salaries of assistants, supplies, and so forth.23 Excluding practice costs from the reform, therefore, approximately halved the redistributional effect.

Although some practice costs can be directly associated with individual services,

many are joint costs, the allocation of which to any specific service is arbitrary (e.g., the rent, the telephone bill). Nonetheless, there was continuing political pressure from the potential winning specialties to include practice costs, and in the Balanced Budget Act (BBA) of 1997 the Congress mandated a three-year transition for the inclusion of “resource-based” practice costs, which has now been accomplished. After an initial unsuccessful try at allocating such costs empirically, the Health Care Financing Administration (HCFA) resorted to physician judgment to allocate practice costs among procedures.

A curious development stemmed from the rhetoric that accompanied the 1989

reform – that by setting a target spending level, physicians would have an incentive to reduce volume in order to increase their fees. Such rhetoric was economic nonsense, since it ignored each physician’s incentive to free ride and increase income by providing more services. Around the time of the reform, however, the rate of increase in the volume of surgical procedures fell substantially for reasons that even today are not entirely clear. Seizing upon the rhetoric, surgeons took credit for this fall and successfully argued that their services should not be pooled with those of other, more profligate physicians. As a result, the Congress in 1990 created a separate performance standard for surgical services effective in 1991, which implied a separate conversion factor (price per relative value unit) for surgical services. Primary care physicians then asked to have a separate target and conversion factor for evaluation and management services and other primary care services, so that the increasing volume of non-surgical procedures such as endoscopy and coronary angiography would not drag down their fee increases. In OBRA93 Congress also granted this request, effective in 1994, so there were then three conversion factors.

By law the increases in the three conversion factors were inversely related to

volume increases for each of the three groupings of services, and the increases differed across the three targets each year. After a few years the conversion factors differed by 21 percent, with surgical services having the highest conversion factor.24 But this spread undid whatever logic there was to the initial linking of procedures across specialties in the relative price scale. I return to issues of physician payment in my discussion of the BBA below.

23 The one-half figure varies considerably by specialty; for example, it tends to be high in radiology because of expensive office-based capital equipment and low in psychiatry. 24 As a result of having low volume increases, surgeons in 1994 and 1995 received 10.0 and 12.2 percent (nominal) increases in their conversion factor, whereas other non-surgical, non-primary care services only received increases slightly over 5 percent in each year.

9

Two Non-Events in the 1990-1996 Period. Two proposals from the 1990-1996 period were not enacted into law but set the stage for future events. The Health Security Act mostly left Medicare alone, reflecting its political sensitivity. As described in the chapter in this volume on Health and Tobacco, however, the Health Security Act did propose an outpatient drug benefit for Medicare. After the Act failed to be enacted, the Administration returned to this issue in its second term, a topic I take up below.

Second, Medicare was a prime target for budget savings for the Republican

Congress that came to office in the November 1994 elections. Among other issues was the elderly’s share of Part B premiums. In 1966, when Medicare was enacted, the elderly paid 50 percent of Part B premiums and the other 50 percent came from general revenues. Part B spending, and hence the elderly’s premium payments, increased more rapidly than the elderly’s income, however, so in 1972 the Congress limited the increase in the elderly’s premiums to the cost-of-living increase in Social Security payments. With this limit in place, the elderly’s share of Part B spending fell to 25 percent over the next decade. At that time the Congress began to regularly pass legislation setting the premium at 25 percent of the cost. In 1990, however, Congress mandated specific dollar increases in premiums rather than a percentage for the 1991-1995 period; the dollar amounts were intended to keep the elderly’s share at about 25 percent. Part B costs, however, grew less rapidly than the Congress had projected, so that the elderly’s share of the premium rose from 25 to 31.5 percent.

In 1995 the Republicans proposed to keep the share at 31.5 percent to achieve

budgetary savings, but the Democrats wished to return to 25 percent. Given the projected increases in Part B spending, even a constant share of 25 percent implied an increase in beneficiary premiums that would be greater than the increase in income among the elderly. This particular Republican proposal was among the most salient issues in the three-week shutdown of the federal government in late 1995, and both it and the shutdown were used to great political advantage by President Clinton in his 1996 re-election campaign.

The Republicans made numerous other proposals for reductions in Medicare

spending at this time. Despite the Administration rejection of these Republican proposals in 1995, it was clear that Medicare could not keep on doing business as it had in the past. The groundwork for the Balanced Budget Act of 1997 had been laid.

The Balanced Budget Act By far the most important piece of legislation affecting Medicare in the 1990s was

the Balanced Budget Act or BBA, enacted in 1997. 1997 was still an era with federal budget deficits projected as far as the eye could see, and the Administration and the Republican Congress agreed on the desirability of reducing the deficits. Part B of Medicare, with its large budget share, could not be left unscathed. Moreover, in 1996 the Trustees of the Medicare Hospital Insurance Fund (“Part A”), who are charged with projecting future spending and Trust Fund balances, had estimated that the Part

10

A Trust Fund would have a zero balance in 2001.25 Outlays had slightly exceeded income in 1995 and were expected to exceed income by ever greater amounts in all subsequent years. And the budgetary picture became much grimmer when the leading edge of the baby boomers started to turn 65 in 2010 and especially when they started to turn 75 in 2020.

The projection of a zero balance in the Part A Trust Fund in 2001 was translated

in the press as Medicare’s “going broke,” the political equivalent of shouting “Fire” in a crowded theater. Thus, the pressure was on both the Congress and the Administration to reduce the rate of growth in Medicare spending, and especially Part A spending, since there was no appetite on either side of the aisle for an increase in payroll taxes to finance Part A.

This pressure resulted in the BBA, which made sweeping changes in how

Medicare paid health care providers. Fundamentally it all but eliminated the considerable portion of cost reimbursement that remained in the program. In doing so, however, it created a new set of problems. I first describe several of the changes made by the BBA and then turn to some of the problems those changes created. I focus especially on post-acute care services, HMO reimbursement, and physician payment.

Post-Acute Services After 1988 the use of post-acute services – home health services, Skilled Nursing

Facility (SNF) services, rehabilitation services, and long-term hospital services – began to rise at very high rates (Table 2).26 From 1988 to 1997 home health visits per beneficiary increased by nearly a factor of 7 and SNF days per beneficiary by nearly a factor of 5. Spending on post-acute services rose from about 3 percent of Part A spending in the mid 1980s to 26 percent of Part A spending in 1996, a time when Part A spending itself was rising at rates well in excess of the growth in tax revenues.27 Spending on post-acute services appeared out of control.

The fundamental causes for the increased spending were several: reimbursement

that was primarily cost-based (though this was a readier explanation for high rather than increased spending); generous reimbursement of new entrants; the incentive the PPS offered to unbundle inpatient hospital services to post-acute sites; and, in the case of home health services, an ill-defined criterion for eligibility for benefits. In addition, court decisions in 1986 and 1988 held that prior HCFA regulations and interpretations of law, which had been used to hold down home health and SNF

25 The trustees are the Secretaries of the Treasury, Labor, and Health and Human Services, and the Social Security Commissioner, and two public trustees, one appointed by Congressional Democrats and one by Congressional Republicans. 26 Long-term hospitals are a heterogeneous group of hospitals with an average length of stay greater than 25 days. Because of their long length of stay, they were exempted from the PPS. To spare the reader, I do not discuss them further, and I ignore issues pertaining to psychiatric hospitals as well. 27 Nominal spending in Part A grew by a factor of 3 between 1984 and 1996, whereas federal tax revenue grew only by a factor of 2.2.

11

spending, were illegal.28 Furthermore, states became more aggressive about shifting some chronic long-term care spending from the Medicaid budget to Medicare, which added to the increase in SNF spending. I now analyze Medicare payment policy for post-acute services in more detail.

Reimbursement for Post-Acute Services Prior to the BBA. I focus on the three

largest components of post-acute spending, home health services, SNF services, and rehabilitation hospitals and units. In 1996 Medicare spent $16.8 billion on home health services, $9.6 billion on SNFs, and $4.6 billion on rehabilitation facilities (Health Care Financing Administration, 1999), (Medicare Payment Advisory Commission, 1999a).

Prior to the BBA, home health agencies were paid their cost per visit up to a limit

of 112 percent of the national mean cost per visit.29 SNFs were paid their cost per day for routine expenses up to a limit of 112 percent of the national mean.30 Ancillary services delivered to SNF patients, for example physical and speech therapy, were reimbursed on a reasonable cost basis, as were capital costs.31 For rehabilitation hospitals and units, reimbursement per admission was related to a target figure, which was a base year cost per admission trended forward by the Consumer Price Index. Hospitals and units shared half of any deviation from this target within a band of 90 to 110 percent. Outside that band the government bore all costs and kept all savings.

New Entrants. New home health agencies and SNFs received cost

reimbursement, just as older entities did. New rehabilitation hospitals and units (units were units within acute care hospitals) were reimbursed their costs for the first three years; the costs in the second full cost reporting period were used to set their target value for future reimbursement. Not only did cost reimbursement give little incentive to economize on initial costs, to the degree that the new rehabilitation units could subsequently economize or reduce per case costs as volume increased, they could keep half of the cost reductions from their target values to a maximum of 10 percent.

The generous reimbursement rules for new entrants caused the number of post-

acute care facilities to rise rapidly. The number of SNFs grew 6.8 percent per year between 1990 and 1996, the number of rehabilitation hospitals and units rose 4.3 percent annually, and the number of home health agencies grew 9.3 percent annually (Prospective Payment Assessment Commission, 1997). Hospitals had an additional

28 The case on home health was Duggan v. Bowen, 691 F. Sup. 1487 (D.D.C. 1988) and on SNF services was Fox v. Bowen, 656 F. Sup. 1236 (D. Conn. 1986). 29 The text oversimplifies in that there was a cost limit of 112 percent for each type of covered service (e.g., physical therapy, home health aide), although the limit was applied to aggregate agency payments. Also, the labor portion of the limit was adjusted by the hospital wage index for the area. 30 Routine services are room, board, and nursing services. This was the rule for freestanding SNFs. Reimbursement for hospital-based SNFs was somewhat greater. 31 Ancillary services at SNFs grew particularly rapidly. In 1990 charges for physical, occupational, speech, and respiratory therapy were 15 percent of total Medicare SNF charges; by 1994 they were 30 percent (U.S. House of Representatives, 1996).

12

incentive to open or acquire these entities, because they could allocate some joint costs to these units (e.g., the CEOs salary), where they would be reimbursed, and away from inpatient services, where reimbursement was fixed by the PPS. Relative to freestanding entities, therefore, the number of hospital-based SNFs and rehabilitation units grew particularly rapidly in the 1990-1996 period, at annual rates of 10.5 percent and 5.8 percent respectively.

Unbundling. The fixed PPS payment per inpatient admission together with the

additional reimbursement for post-acute services offered hospitals an incentive to unbundle inpatient services, that is, to substitute post-acute care services for the last days of stay in the hospital. As a result, length of stay and beneficiary days per thousand fell, and post-acute services rose (Table 3). That much of this fall could be attributed to the payment system can be seen by comparing the fall in length of stay between 1988 and 1996 for Medicare beneficiaries and all others, whose inpatient stays were usually paid on a per day rather than a per admission basis. Medicare length of stay fell 27 percent; length of stay for all patients (including Medicare) fell 15 percent.32 Moreover, among the ten DRGs with the largest number of post-acute care users, average length of stay dropped 1.3 to 2.0 days between 1994 and 1996 among post-acute care users, but only 0.6 to 1.8 days among non-users (Medicare Payment Advisory Commission, 1998a).33 Finally, hospitals that operated post-acute services had greater drops in length of stay (Prospective Payment Assessment Commission, 1996).

At the time of the BBA hospitals were enjoying the highest margins the industry

had experienced since the first two years of the PPS in 1984 and 1985 (Table 3).34 The Medicare Payment Advisory Commission (MedPAC) estimates that unbundling was responsible for a substantial portion of this fiscal improvement by lowering hospital costs per Medicare (inpatient) case by about 10 percent with no corresponding adjustment in payment (Medicare Payment Advisory Commission, 2001a).

The BBA Provisions for Post-Acute Care. The BBA addressed the unintended

overpayment to hospitals in five ways. First, in an effort to better match payment for inpatient services with the cost reductions from unbundling, which had not been accounted for in setting prior hospital updates, the hospital update factor for 1998 was zero. Further, the BBA mandated that future updates to the PPS be below the rate of increase in hospital input prices. MedPAC estimates that these reductions have to date accounted for about two-thirds of the overpayment from unbundling; in other words, the payment rate for inpatient services is still about 3 percentage points above

32 Although this comparison does not control for any age-specific factors affecting length of stay (it is not clear which direction they would go), it probably understates the contribution of the payment system because several private payers changed from paying hospitals a daily rate to a DRG basis of payment in this period. 33 And the drop for users was greater than for non-users in each of the ten DRGs. 34 The PPS was supposed to have been introduced on a budget neutral basis, but was misnormed, so that initial payments were too great.

13

where it would have been if the rate had been adjusted for the unbundling – and of course hospitals have profited from the unbundling in the interim.

Second, the BBA attempted to end cost reimbursement for post-acute services by

mandating that HCFA develop prospective payment systems for those services. In particular, HCFA was to put in place prospective payment systems for SNFs in 1998, for home health in 1999, for rehabilitation services in 2000, and was to develop (but not implement) a system for long-term hospitals in 1999.

These mandates to develop new payment systems created an enormous workload

for HCFA. The very short time frames reflected the Congress’ frustration with the rise in spending and HCFAs inability to introduce prospective systems for post-acute payment despite prior requests. HCFA, however, faced the reality that in contrast to the PPS for inpatient services, there was little prior experience with prospective payment systems for post acute facilities, and that the data available to the agency for developing the systems were sparser than the data for inpatient care at the time the hospital PPS was developed.

The agency nonetheless met its first target by introducing the Resource Utilization

Group (RUG) system for SNFs in July 1998. This system had been developed for chronic long-term care services, and when transplanted to the post-acute environment it proved to have problems. Nonetheless, this was probably the only option open to HCFA if the schedule mandated in the BBA was to be met. In October 2000 the Agency introduced a prospective case-based system for home health agencies, the Home Health Resource Groups (HHRGs), and in April 2001 it introduced a prospective system for rehabilitation hospitals and units.35 Despite the problems with reimbursement for post-acute care that I describe below, HCFAs performance in introducing these systems in the space of a few years, largely on the timetable mandated in the BBA, was commendable.

Third, just as it had with inpatient hospital services, the BBA reduced the growth

rate of reimbursement for post-acute services substantially. In the case of SNFs, these reductions came through the PPS, which ended cost reimbursement and gave the Congress control over rate increases. In the case of home health care, HCFA was given two years to develop a prospective payment system. In the interim the BBA mandated a payment system that substantially tightened the limits on cost reimbursement. Previously the limit on cost reimbursement was 112 percent of the mean cost of freestanding agencies; this was reduced to the lesser of 105 percent of the median cost or a blend of the agency’s 1994 cost and the average cost in the Census region.36 Because of the increase in cost after 1994, the blend was usually the binding constraint. Moreover, the BBA provided that when the prospective payment system was ultimately introduced, there was to be a 15 percent reduction from a

35 The rehabilitation system was called the Functional Independence Measure/Functional Related Group system (FIM/FRG). 36 The blend was 98 percent of a weighted average of 75 percent of the agency’s 1994 cost and 25 percent of the average cost per beneficiary in the census region.

14

budget neutral amount. The Congress clearly considered home health spending excessive and was intent on rolling back spending to earlier levels.

Fourth, the BBA substantially tightened the rules for new entrants. As of 1998

target amounts for new rehabilitation facilities were limited to 110 percent of the national mean target amount even if their initial costs were greater. New SNFs were subject to the prospective payment system for SNFs that was implemented in 1998, and new home health agencies were subject to the revised payment limits in the interim system.

Fifth, the BBA modified the DRG payment for patients in certain DRGs who

were discharged after a relatively short stay and who used post-acute services. In particular, HCFA was to specify ten DRGs; for patients in those DRGs who were discharged before the geometric mean length of stay for that DRG, the hospital would be paid per day rather than per admission. In implementing this regulation HCFA chose ten DRGs with high use of post-acute services. For those ten DRGs this change made the payment system more neutral (lower powered in contract theory jargon) with respect to whether the marginal day would be spent as an inpatient or in a post-acute setting.

In addition to these changes, the BBA attempted to clarify eligibility rules for

home health services. Patients are eligible for Medicare home health benefits even without a hospital stay if they are homebound and need part time or intermittent skilled nursing services or physical or speech therapy.37 These terms had, however, not been defined in statute. After the 1988 court decision substantial ambiguity about the meaning of the terms remained; the BBA defined these terms.38

Under the old definition and because no hospital stay was required, a number of

home health services that were essentially chronic long-term care services rather than post-acute services had de facto become covered Medicare services. For example, in 1994 over half of all home health visits went to the 12 percent of users who had more than 150 visits per year. This group averaged 275 visits per user per year, or nearly daily visits (Prospective Payment Assessment Commission, 1997). This group received a disproportionately high percentage of nurse aide as opposed to nurse visits; such visits are more likely to be of a chronic nature; the aide, for example, might give

37 Prior to 1980 beneficiaries needed to have a hospital stay, but that requirement was dropped on the argument that making home health services more freely available would pay for itself by reducing nursing home costs, an argument which later data has not supported (Kemper, 1988). The lack of a hospital stay has led to a very heterogeneous population of users, some of whom are truly receiving post-acute services (e.g., a few visits to verify that recuperation is normal), others of whom are receiving almost daily visits indefinitely because they can barely manage to live independently. The 1988 court decision referred to above were directed at earlier HCFAs regulations that required patients to need part time and intermittent services rather than part time or intermittent services. 38 For the purpose of establishing initial eligibility intermittent care is care that was needed (strictly) less than 7 days a week or less than 8 hours a day for 21 days or less. Once eligible, however, an individual can receive services for any number of days per week, as long as the total is less than 8 hours per day and 28 hours per week.

15

an elderly person assistance with bathing. I doubt that the BBAs definitions of eligibility will much reduce this kind of service, although the new prospective payment system may well do so.

One change the BBA made was mainly an accounting change. The BBA shifted

the cost of all home health visits over one hundred in a year plus all visits that did not follow a hospital stay from Part A to Part B. Although this change did not affect payment rates by beneficiaries (the BBA waived the Part B coinsurance on such services) nor payment rates to home health agencies, both the Administration and the Congressional Republicans could take credit for prolonging the life of the Part A Trust Fund. The only real effect, however, was to shift the cost of these services from payroll tax to general revenue financing.

Shifting State Monies to Medicare. Another cause of increased post-acute care

spending was left untouched by the BBA. Patients are eligible for the SNF benefit if they stay in the hospital at least three days.39 The SNF benefit then covers up to 20 days of a stay in the facility with no copayment and another 80 days with a copayment of one-eighth the hospital deductible (in 2001 this is just under $100). Although intended to apply to patients who need a period to recuperate from their illness but do not need the intensity of service that a hospital provides, states have aggressively pursued Medicare reimbursement for Medicaid-eligible nursing home residents who are hospitalized and then return to the nursing home, which will typically have a SNF facility. New York for example, enacted a statute that Medicare rather than Medicaid would always be billed first to determine if it would pay. If Medicare rather than Medicaid pays for any of the nursing home stay, of course, the states save their share of Medicaid reimbursement of the nursing home. Medicare’s responsibility in principle ends when the episode of illness that caused the hospitalization ends (up to a maximum of 100 days of SNF care), but defining the end of the episode is often ambiguous, especially in the nursing home population.

As a result of state efforts to shift the financing of nursing home residents from

Medicaid to Medicare, the proportion of nursing home revenues that come from Medicare has risen substantially. Only 2 percent of nursing home revenues came from Medicare in 1980 and even as late as 1991 only 3 percent came from Medicare. By 1998, however, this proportion had risen to 12 percent, another example of chronic long-term care services being covered by Medicare (Cowan, et al., 1999). Whether this trend has run its course or not is unclear.

The BBA Provisions for Medicare+Choice During the 1990s enrollment in Medicare HMOs grew from 3 percent of

beneficiaries to about 15 percent (Figure 2). This growth was prompted in large part

39 The three-day stay requirement for SNF services has been contentious at least since the same requirement was removed for home health services around 1980. The Catastrophic Coverage Act of 1988 abolished the three-day stay requirement for SNF services, and SNF usage shot up in 1989 (see Table 2). When the Act was repealed in 1989, the three-day stay requirement was reinstituted.

16

by inefficiencies in traditional Medicare that were then transmitted to HMO reimbursement through the AAPCC. (Recall that the AAPCC was set at 95 percent of traditional Medicare payment in a county.) HMOs could potentially exploit these inefficiencies. HMOs were concentrated in larger metropolitan areas, where there were often several HMOs, and competition among them forced the efficiency gains through to beneficiaries in the form of additional benefits and reductions in premiums. Table 4 shows that the estimated value of these benefits was directly related to the level of plan payment, averaging $121for the decile of plans in the counties with the highest reimbursement rate, which was 29 percent above the national average, but only $48 in the decile of plans with the lowest reimbursement rates.40

The growth in HMO enrollment posed two issues. First, the geographic

dispersion in reimbursement became much more visible. As long as Medicare simply sent checks to individual physicians and hospitals on a service-by-service basis, the geographic dispersion in payments, which was well known to researchers, seemed to remain below the political radar screen. When it was summarized by the AAPCC, however, it did not. And the dispersion was large; Table 5 shows reimbursement in the six highest and six lowest counties in the country in 1997. The dispersion is more than a factor of three, so it cannot be explained by variation in factor prices; the spread between the minimum and maximum wage rate index is a factor of two, and several inputs are purchased in national markets with the same factor price.41 This geographic variation mirrors findings in the literature that rates of various procedures in the Medicare population vary among states or large sub-state areas by a factor of 4 (Chassin, et al., 1986).

Politically this dispersion created a demand for more equality, especially from

rural areas, which are well represented on the Senate Finance Committee. But the demand also came from metropolitan areas such as the Twin Cities and Portland, Oregon, which had AAPCCs that were only around 50 to 60 percent as large as those of Miami and New York City. The Congressional representatives from these low rate areas felt that their constituents were not being treated fairly by a program that was in principle to provide uniform benefits across the nation, and of course they were lobbied by local providers to increase rates. After all, why should Medicare beneficiaries in Miami have the option of obtaining drug benefits for joining an HMO, but beneficiaries in Minneapolis not? Indeed, the argument went on, beneficiaries in Minneapolis were being penalized because their physicians were efficient in the ordering of services.

The second issue posed by the growth of HMOs was favorable selection into the

program, so that instead of saving money by creating an HMO option, the government was paying more for those beneficiaries that joined an HMO than it

40 The dollar values in Table 4 almost certainly understate the spread in the value of the benefits, because they have been standardized by the area wage index. The major supplementary benefit, however, is some drug coverage, and drug prices undoubtedly vary less from area to area than the wage index. 41 The fraction purchased in national markets varies by provider type, but is around 30 percent for hospitals.

17

would have paid for them had they remained in traditional Medicare. Estimating just how much more the government was paying was controversial, but the most widely cited estimates were in the 6 to 8 percent range, even net of the 5 percent that the government took off the top (Brown, et al., 1993), (Riley, et al., 1996), (Congressional Budget Office, 1997).

The studies that arrived at these amounts used several methodologies. A popular

methodology was comparing the use of those who enrolled in HMOs in the months before they enrolled, a time when they were enrolled in traditional Medicare, with the use of those who remained in traditional Medicare. Analysts also compared the use of those who disenrolled from HMOs and went back to traditional Medicare in the immediate post-disenrollment period with those who had never enrolled in HMOs.42

An influential study of this sort by the Physician Payment Review Commission

found that those who enrolled in HMOs in the 1989-1994 period spent 38 percent less in the six months before they joined than the control group who did not join, after adjusting for age, sex, institutional status, welfare status, and county of residence. By contrast, those disenrolling spent 42 percent more than the control group (Physician Payment Review Commission, 1996). In other words, the healthy appeared to be joining HMOs and the sick appeared to be leaving them. The same study compared mortality rates of those who joined HMOs in the year following enrollment with a control group who never joined HMOs, controlling for age, sex, and county of residence. The mortality rate among those who joined was 25 percent less than in traditional Medicare.43 Although an HMO lobbyist might claim some health benefits from the more complete and integrated care that HMOs could in principle provide, it would be hard to maintain that an HMO could cause anything like a 25 percent reduction in mortality and keep a straight face. The fact of selection was further confirmed in another study that compared the self-rated health status in 1994 of those in HMOs and those in traditional Medicare; beneficiaries in HMOs reported themselves in better health (Riley, et al., 1996).

The HMO industry’s response to these findings was to assert that they were based

on data from a period when HMO enrollment was only a tiny share of Medicare, but by 1997 the share was much higher and therefore it was likely that HMOs had a more representative risk mix. Moreover, HMOs asserted that it gave a misleading picture to use data from the period around the time of enrollment because as HMO beneficiaries remained in the HMO they would age and therefore their spending and mortality experience would regress toward the mean. This latter argument, of course, did not deny that some overpayment existed around the time of enrollment.

42 One reason for this methodology (rather than, for example, comparing use of the same beneficiary before and after joining an HMO) was that reporting of use by HMO enrollees was incomplete. Although HMOs were in principle to report use by filling out dummy claims forms, HCFA was not able to enforce this rule completely. The BBA strengthened HCFAs hand considerably in this regard, and reporting is now much better. 43 As is well known, spending at the end of life is costly; in a given year between 25 and 30 percent of Medicare spending is on behalf of the 6 percent of beneficiaries who die (Lubitz and Riley, 1993). With disproportionately low mortality rates, HMOs profit by avoiding this end-of-life spending.

18

In response to these arguments, the BBA mandated a study by MedPAC similar in

method to the Physician Payment Review Commission study of selection discussed above, but one that used more recent data. Although this subsequent study had some methodological differences with the earlier study, it essentially confirmed its findings. In particular, it confirmed that those enrolling in 1997 spent 28 percent less in the twelve months before enrollment than those in traditional Medicare (Medicare Payment Advisory Commission, 2000a). (It found no difference in spending between those disenrolling and the control group, however.) Moreover, there was a 21 percent mortality difference in 1998 in the first year after enrollment. Although this mortality difference fell for beneficiaries who had been in the HMO longer, it never returned to the mean, and over all HMO enrollees the mortality difference was 15 percent (Figure 3).

The observed selection could stem from deliberate actions of HMOs such as

marketing to better risks. Or it could occur even if HMOs were completely passive; those who are sick may simply prefer the unmanaged traditional Medicare benefit. Because Medicare beneficiaries can enroll or disenroll in HMOs monthly, it is not difficult for beneficiaries to change plans.44 Whether or not the selection stemmed from deliberate actions of HMOs, however, was irrelevant to the greater payments that Medicare was making for those enrolled in HMOs than if they had been enrolled in traditional Medicare.

Moreover, the selection was threatening to create ever worse overpayment

through a death spiral. If those joining HMOs in any time period are drawn from the population of the best risks remaining in traditional Medicare, as seems likely, mean costs for the groups in traditional Medicare and in HMOs will both rise but will steadily diverge (Cutler and Reber, 1998).45 That is, the traditional Medicare mean will increase by a greater amount than the HMO mean, so that payment in traditional Medicare, the basis for the AAPCC, will become ever more overstated as a measure of the cost of treating HMO enrollees. In 1997 HMO enrollment in some counties was approaching 50 percent of beneficiaries; thus, the bias could be appreciable.

In addition to addressing geographic differentials in payment and overpayment

from selection, many Congressional Republicans wished to increase enrollment in HMOs.46 But HMOs had several complaints about how Medicare paid them, which they argued reduced their willingness to participate. First, payment could change substantially from year to year, especially in rural areas, for reasons that had little to do with HMO costs (Table 6). Second, within metropolitan areas payment rates could differ substantially by county (Table 7). Just because a person moved her residence across the line from the District of Columbia to Montgomery County, for

44 In general, however, those joining an HMO give up their right to obtain an individual Medigap plan without medical underwriting should they subsequently elect to disenroll and move back to traditional Medicare. 45 This result uses the fact that the distribution of health care spending is approximately lognormal. 46 And many in both parties saw an HMO option as a way to provide drug benefits to their constituents.

19

example, did not justify a thousand dollars less annual payment for that person’s care. Finally, physicians lobbied for the rights to form their own HMOs, since they wished to appropriate monies that were otherwise going to health plans.

The BBA and HMOs. The BBA responded to these various complaints and ills in

several ways. To address the differences in payment rates by area, it created a floor of $367 per month ($4,404 per year) for the AAPCC in 1998. This floor was binding in counties with about 12 percent of beneficiaries, or about half of all non-metropolitan beneficiaries. It also attempted to move the AAPCC partway toward a national payment rate by creating a blend of the county rate and the national mean rate. For each of the next five years, the weight on the national rate was to increase by 10 percentage points, so that by 2003 there would be a 50-50 blended rate. In order to protect the high payment areas, however, all HMOs would get at least a 2 percent annual update. Thus, payment to an HMO was to be the maximum of the floor payment, the blended rate, or the 1997 rate updated by 2 percent per year.

In the Benefits Improvement and Protection Act (BIPA) of 2000 the Congress

went further along this road, increasing the floor payment in non-metropolitan areas by 18 percent, from $402 per month in 2000 to $475 in 2001, and introducing a even higher floor of $525 $(6,300 per year) for metropolitan areas with more than 250,000 population. These increases make the floor rates binding for 40 percent of the beneficiaries.

In response to selection, the BBA took two actions. First, it lengthened the lock-

in period for enrollees who opted for HMOs, although implementation of this provision was deferred for five years. Previously, enrollees had been allowed to opt into or out of health plans on the first of every month, unlike the private sector where enrollees are typically locked into their plan for a year. Although enacted as a beneficiary protection, the ability to change plans monthly clearly increased the opportunity for selection. The BBA changed the monthly lock-in to an annual lock-in period starting in 2002, with the proviso that enrollees could disenroll monthly for the first three months of the period.47

Second, the BBA mandated that HCFA introduce a risk adjustment method based

on health status by 2000. In other words, in addition to the age and sex of the beneficiary, HCFA was to account for the health status of the beneficiary in determining how much to pay the HMO. On the correct assumption that HCFA would want to use a method that accounted for a beneficiary’s diagnosis, the BBA mandated that health plans report information on diagnosis to HCFA. Many health plans only had accurate diagnostic information on those who had been hospitalized, so the BBA required that initially only inpatient diagnostic information be reported, but it also set a requirement that outpatient information ultimately be reported. Because it would partially correct for selection, risk adjustment based on health status would reduce payment for the average health plan; as a result, the BBA provided for a transition to fully risk-adjusted payment.

47 In 2002 they could disenroll for the first six months of the period.

20

In response to the volatility of rates, the BBA fixed in statute annual updates to

the AAPCC as a percentage increment from the prior year’s rate, so that updates were no longer a function of the past year’s spending in traditional Medicare. This also eliminated the possibility of ever increasing overpayment at the national level or a death spiral, since annual updates in payment to HMOs and to traditional Medicare could differ. In response to physicians’ wishes to form HMOs, the BBA permitted this, but there has been minimal response to this provision.

Two New Options. The BBA also created two new options for beneficiaries,

Medical Savings Accounts (MSAs) and a private fee-for-service option; the ensemble of these options, plus the HMO option, were now termed Medicare+Choice to distinguish them from traditional Medicare. MSAs are plans with a large deductible (the legislation set this between $3,000 and $6,000 per person annually). If a beneficiary elected an MSA, he or she would receive the AAPCC, adjusted by the relevant risk adjusters, less a premium for the large deductible plan. This amount would not be taxable, provided it was spent for medical care. Unspent balances could accrue over time and could ultimately be used to finance long-term care or used as a bequest.48

There was intense political activity around MSAs. Several Republicans saw

MSAs as a way to reduce moral hazard; most Congressional Democrats and the Administration saw them as a further opportunity for selection, because only good risks would profit from opting for a plan with a large deductible. Because of inadequate risk adjustment methods, such selection would benefit well off and healthy seniors at the expense of the Treasury. The compromise was an agreement for a demonstration project that would be capped at 300,000 participants. Subsequently, however, no insurance company entered the market offering MSAs, so the issue was moot.

The private-fee-for service option stemmed from complaints by physicians that

Medicare’s limitations on balance billing (billing above the Medicare fee schedule) abridged their freedom and the freedom of the beneficiary to contract for a fee. This had been a longstanding issue with physicians, who, at the time of Medicare’s enactment, had often price discriminated among patients according to ability to pay (Kessel, 1958). When it began, therefore, Medicare offered weak incentives to not balance bill, but in the 1980s these limitations were greatly strengthened, and physicians were not permitted to bill more than 10 percent above the fee schedule. Particularly in the light of the changes in relative fees that were being introduced, physicians lobbied for “freedom to contract.”

The religious right also supported the private fee-for-service option. They were

concerned that increasing fiscal constraints in the Medicare program would lead to restrictions on care near the end of life. They therefore supported structuring the private fee-for-service option so that physicians must be paid on a fee-for-service

48 If used as a bequest, it would be taxable.

21

basis (i.e., no element of capitation) and their fees could not be reduced as a function of utilization (i.e., no bonus to the physician for keeping utilization low).

Thus, beneficiaries electing the private fee-for-service option buy an insurance

policy that insures Medicare covered services. The insurer is allowed to contract with physicians at any rate mutually agreed upon and can charge beneficiaries a premium to cover any increment in fees above the Medicare fee schedule. To help pay the premium for this policy, the beneficiary receives the AAPCC. Unlike the MSA option, the private fee-for-service option appears as if it will have considerable consequences, as I come to below.

The BBA and Physician Payment At the time of the BBA, payment to physicians suffered from two problems, one

related to the effort to cap physician spending and one related to relative prices. First, as described above, under the VPS increases in the target spending for physician services were in part a function of a five-year average of the rate of increase in the number of units of services that were delivered. The five-year average was conceived of as a measure of the cost of technological change. But the volume of surgical procedures had stopped increasing very much. Because their spending target was based on much higher historical volume, in the short run the fall in volume led to very large increases in surgeons’ fees, 10 percent in 1994 and 12.2 percent in 1995.49 In the longer run, however, the low annual rates of increase in volume started to markedly reduce the five-year average rate of increase, and by the time of the BBA, it had affected the average enough that, together with other actions the Congress had instituted to reduce spending on physician services, nominal physician fees were projected to fall for at least the next ten years.50 This was not tolerable politically, in part because fees in the commercial market were not expected to fall in nominal terms, and a substantial divergence between Medicare and commercial fees might lead some physicians to refuse to see Medicare patients.

The Congress therefore abandoned the VPS and substituted the Sustainable

Growth Rate (SGR) system. Under this method, increases in target spending are a function of real GDP growth rather than a function of past increases in units of service. The rationale was that physician payment should be tied to the government’s ability to pay. Because GDP growth was very robust in the immediate post-BBA years, physicians did well. I discuss below whether a spending cap for physician services is good policy.

There was also a problem with relative prices. As described above, the Congress

had established three conversion factors, which were diverging, thereby undermining whatever rationale the relative value scale had. In the BBA the Congress therefore

49 Inflation in those years was about 2 percent, so these were large real increases by historical standards. 50 The most important of the other actions that the Congress had instituted to reduce spending on physician services were arbitrary continuing reductions in the target. These reductions had reached 4 percentage points per year by the time of the BBA.

22

returned to a single conversion factor. Because this was done in a budget-neutral fashion, the conversion factor for surgical procedures was reduced a little more than 10 percent. Hence, surgeons took a double hit; the conversion factor for surgical services was reduced, and so-called resource-based practice costs were phased in, as described above. Both these changes have now been accomplished, and, as far as is known, most surgeons continue to accept Medicare patients, suggesting that there were rents in surgical fees.

Other Changes Made by the BBA The BBA made a host of other changes in Medicare, four of which I briefly

describe here. These four related to hospital outpatient departments, teaching hospitals, disproportionate share payments, and additional preventive care benefits.

Hospital Outpatient Departments. If a physician treats a patient in the outpatient

department or in an ambulatory surgery center rather than in an office, Medicare reimbursement is complicated. The physician receives payment under the physician fee schedule, but for 650 services there is a “site-of-service” adjustment, which is a 50 percent reduction in payment for practice cost on the grounds that Medicare is also paying the hospital a “facility fee” to the outpatient department or ambulatory surgery center to cover the costs of personnel and supplies. Between 1984, the beginning of the PPS, and 1996 these facility fees grew by more than 12 percent per year and amounted to $10.5 billion in 1996.51 Both the growth and the cost-based nature of the facility fees led the Congress to mandate that HCFA develop a prospective payment system for hospital outpatient departments and ambulatory surgery centers. I turn below to the problems of this system.

A second provision applying to outpatient departments concerned beneficiary cost

sharing. Due to a defect in the way the law had been written in the late 1980s, hospital outpatient departments were able to increase revenues from patients by raising charges, but this had the effect of raising the cost sharing percentage for beneficiaries to an average around 50 percent, far above the standard 20 percent figure for Part B (Medicare Payment Advisory Commission, 1998b).52 Moreover, the cost sharing percentage varied substantially by service (Table 8). For some outpatient department services the cost sharing was at or near the statutory 20 percent for other Part B services; for others, it was in the 60 to 80 percent range.

The BBA sought to remedy this, but to correct it fully (i.e., to bring all services

back to a 20 percent coinsurance rate) was estimated to cost $4 billion in an Act that

51 This figure is understated because it does not include HMO enrollees, the fraction of which grew substantially over this period. 52 Medicare continued to base its reimbursement to outpatient departments on 20 percent of their costs, but hospitals were permitted to charge beneficiaries 20 percent of their charges. At the time this law was enacted, charges and costs were not much different, but over time they differed sharply as hospitals raised charges.

23

was trying to reduce Medicare expenditure.53 The BBA therefore put in a very lengthy transition to reduce beneficiary coinsurance. It held that nominal coinsurance amounts for each service would be fixed at 1999 levels until they reached 20 percent, something that was estimated could take up to 40 years for some services.

Teaching Hospitals. Because they had higher costs per admission, even

controlling for DRG, teaching hospitals had been paid more per admission than non-teaching hospitals since the inception of the PPS in FY1984. The amount of the increment was a function of the house-staff-to-bed ratio. (House staff are interns and residents.) The initial increment that the Department of Health and Human Services (DHHS) proposed came from a regression of the logarithm of cost per case at each hospital on the logarithm of (1 plus the house-staff-to-bed ratio), the logarithm of the hospital’s case-mix index, the logarithm of the area wage index where the hospital was located, and, importantly, the logarithm of the number of beds at the hospital, which was positively correlated with per case costs.54 The increment the Department proposed was based on the coefficient of the house-staff-to-bed ratio, which was strongly correlated with cost per case. Paying on the basis of the house-staff-to-bed ratio, however, offered teaching hospitals a subsidy for increasing the numbers of house staff (residents).

Moreover, the payment formula in the PPS did not include the number of beds on

the grounds that hospitals should not be offered an incentive to increase the number of beds. Teaching hospitals tend to be large (i.e., the number of beds and the house-staff-to-bed ratio are positively correlated); as a result, if the regression had mimicked the payment formula, the coefficient on the house-staff-to-bed ratio would have been higher. Teaching hospitals rightfully complained that they would be damaged by the formula the Department proposed. Congress was in a rush to pass legislation authorizing the PPS and, in an effort to appease the teaching hospitals, simply doubled the increment that the DHHS proposed, writing into statute twice the regression coefficient, or an 11.59 percent increase in payment for each 0.1 increment in the house-staff-to-bed ratio. This must be one of the few or perhaps the only instance of Congress writing into legislation a regression coefficient to four significant digits. The doubling of the coefficient, however, meant payment to teaching hospitals was considerably above the empirical relationship between house staff per bed and cost, even from a regression equation that did mimic the payment formula, so teaching hospitals began to receive a subsidy that had not been present under the prior cost reimbursement regime. Moreover, because the PPS as a whole was to be budget neutral, the subsidy for teaching hospitals was financed by lower reimbursement for non-teaching hospitals.

53 Moreover, the Congressional intent was to reduce the burden on beneficiaries, but over a third of them had employer-provided retiree health insurance that paid much or all of the outpatient department cost sharing (Table 1). The incidence of reducing the coinsurance rate to 20 percent, therefore, would partly redound to shareholders. 54 Dummy variables for city size groups were also included in the initial regression. See (Pettengill and Vertrees, 1982).

24

Doubling the coefficient also doubled the financial incentive to hire more residents. Consistent with this incentive, the stock of residents increased by about a third from 1985 to 1993, after which it stabilized, suggesting teaching hospitals had adjusted to the new financial regime (Table 9). Because the output of US medical schools has been approximately constant since the mid 1970s, this increase came from hiring more residents from non-US medical schools and from lengthening training periods. The increase in residents occurred at a time when many felt the United States had at least an adequate supply of physicians. And the additional payment to teaching hospitals, of course, increased approximately proportionately to the increase in residents, reaching $7 billion by the mid 1990s, whereas in 1985 it was only $1.4 billion (Newhouse and Wilensky, 2001).55

Preserving the additional payments to teaching hospitals was especially important

to Senator Moynihan (D-NY), the chairman of the Senate Finance Committee from 1992-1994 and the ranking minority member after 1994 until his retirement in 2000, but the teaching hospital payments attracted bipartisan support.56 As a result, the Congress did not eliminate the subsidy to teaching hospitals, but after 1984 it steadily chipped away at it in several budget bills.

By the time of the BBA the payment had been reduced from an 11.59 percent

increment for every 0.1 increment in the house-staff-to-bed ratio to a 7.5 percent increment.57 The BBA further reduced this amount by 0.5 percentage points per year for four years down to a 5.5 percent increment. The Balanced Budget Refinement Act of 1999 and the BIPA legislation of 2000 have subsequently stopped the transition at 6.5 percent, a figure now scheduled to drop to 5.5 percent for 2003 and beyond. Even the 5.5 percent figure, however, represents a subsidy to teaching hospitals.58

The BBA also capped the number of residents that a hospital could count toward

reimbursement at 1996 levels. Thus, the hospital could no longer gain additional revenue by expanding its number of residents, but it could still lose monies by contracting them.59 Because of the stability in numbers of residents since 1993, however, the cap at 1996 levels is not very binding. The problem remains of how to pay teaching hospitals without distorting the market for house staff.

55 The amounts include both Direct and Indirect Medical Education payments. 56 Although New York received a disproportionate share of the teaching hospital monies, Senator Moynihan’s concern for teaching hospitals went well beyond such parochial interests. He viewed these monies as critical for the advances in knowledge being generated at teaching hospitals. During the debate over the Health Security Act he made a speech on the Senate floor in which he said that to reduce these payments would be “a sin against the Holy Ghost.” 57 Thus, if a 500 bed teaching hospital raised its number of residents from 500 to 550, its payments per case would rise 7.5 percent. 58 The so-called empirical level is now around a 3 percent increment for each 0.1 increase in the ratio. 59 The BBA did offer some very limited transitional funds to hospitals that reduced residents in that payments were reduced over a five year period.

25

Disproportionate Share Payments. In the mid-1980s the Congress enacted a Disproportionate Share Hospital (DSH) program. This program paid hospitals that treated many Medicaid patients more per DRG. Initially this program had the rationale that poor Medicare patients were more expensive to treat, a finding based on an analysis of data from the state of Massachusetts (Epstein, et al., 1990). A later study using national data was able to replicate the finding for Massachusetts, but showed that the Massachusetts finding did not generalize; there was no difference in the cost of poor and non-poor Medicare patients using national data (Kominski and Long, 1997).

Not surprisingly, recipients of DSH monies had an interest in continuing to

receive them, and so the program acquired other rationales. Many legislators, especially Democrats, wanted to use Medicare monies to aid safety-net hospitals with their burden of uncompensated care for the under 65 population. Such safety-net hospitals were usually large urban public hospitals.60 Aid for safety net hospitals would have been more logically financed from general revenues, but it did not appear likely that general revenues would be appropriated for this purpose. Another rationale, more specific to Medicare, was that the safety-net hospitals were important sources of care for those Medicare beneficiaries that lived near them.

At the time of the BBA the Medicare DSH monies amounted to $4.5 billion, up

from $1.1 billion in 1989. Given the various articulated and unarticulated goals for the DSH program, the formula for allocating the monies among hospitals had several important defects. The formula set a threshold before a hospital was eligible for any funds based on the proportion of its total admissions (including the non-elderly) that were eligible for Medicaid and its share of Medicare admissions that were Medicaid eligible; these two proportions were equally weighted. Importantly the formula did not include a measure of uncompensated care or the hospital’s share of patients who lacked insurance coverage. This was a critical omission to the degree that the program was intended to support safety-net hospitals, because the correlation between Medicaid and uncompensated care admissions was weak. Fundamentally the low correlation arose because Medicaid patients were insured; hence, other things equal, the more generous eligibility was for the state’s Medicaid program, the fewer uninsured persons there were in the state. This low correlation was exacerbated in the mid-1990s when certain states began to dramatically expand Medicaid eligibility, most notably Tennessee and Oregon, which expanded these state’s share of the DSH monies.

DSH payments to hospitals were structured as a percentage increment to their

DRG rate, but the increment was sharply progressive in the percentage of Medicaid patients, reflecting the desire to target the urban safety-net hospitals. These hospitals not only had the largest uncompensated care burdens; they had a relatively small share of Medicare patients, and DSH monies were simply additional reimbursement for each Medicare patient. Making the formula progressive, therefore, was a way to

60 Examples would include the Los Angeles County Hospital, the Cook County Hospital, the Boston City Hospital, and the hospitals of the New York City Health and Hospitals Corporation.

26

give the safety-net hospitals more money in spite of their relatively small number of Medicare admissions. Rural hospitals, however, had many Medicare patients (often they were they only local hospital in a community with a high proportion of elderly persons), and the Congress did not intend the progressive formula to apply to rural hospitals. The Congress therefore put in place a much higher threshold for rural than for urban hospitals before they became eligible for any DSH monies. The threshold was so much higher, however, that 96 percent of the DSH monies went to urban hospitals.61 Not surprisingly, rural hospitals lobbied against this unequal treatment.

Finally, there was a notch at the threshold where DSH payments began; urban

hospitals at the threshold received 2.5 percent more for each Medicare admission, whereas hospitals just below the threshold received no increment.

Simply to save money the BBA reduced the funds available for the DSH program

by one percent a year for five years (a cumulative 15 percent at the end of five years) and asked for studies to lay the groundwork for later reform of the DSH program. Most importantly, it mandated that hospitals start to report the share of admissions from uninsured patients, so a measure of uncompensated care could be included in the formula for allocating DSH monies. Subsequently the Balanced Budget Refinement Act (BBRA) of 1999 and the Benefits Improvement and Protection Act (BIPA) of 2000 have mostly restored the BBA cuts in the DSH program; under current law there is only a 2 percent cut in 2001, 3 percent in 2002, and no reductions at all after 2003.

Preventive Care Benefits. The BBA added several preventive benefits to

Medicare. For the elderly, annual rather than bi-annual mammograms were covered.62 Coverage for Pap smears went from every three years to annual. Coverage for prostate cancer screening was instituted for the first time. A number of other preventive benefits were added as well. These were relatively low cost benefits, but they did help generate beneficiary support to offset the resistance from providers for the substantial reductions in reimbursement and helped keep the BBA from being portrayed as simply a takeaway.63

The Effects of the BBA and Some Givebacks The BBA succeeded in its main goal of reducing the rate of growth in Medicare

outlays. Nominal spending in 1998 and 1999 was below spending in 1997, which had never happened before, and in real terms spending in 2000 was still below the 1997 level (Table 10). The reductions were especially great in post-acute care. Between 1997 and 1999 home health spending fell 45 percent, from $17.5 billion to $9.7 billion, and SNF payments fell 17 percent, from $11.0 billion to 9.4 billion (these

61 Although rural hospitals did not receive their proportionate share of DSH funds, there are many special provisions in the PPS to aid or subsidize rural hospitals, including the Critical Access Hospital program, the Sole Community Hospital program, the Medicare Dependent Hospital program, and the Rural Referral Center Hospital program. 62 There was more frequent coverage for those under 65 eligible for Medicare as well. 63 I have been unable to find a cost estimate for these benefits, but my recollection is that they cost around $6 billion over a five year period.

27

figures are nominal dollars). The number of home health users per beneficiary fell more than 20 percent, and SNF discharges declined over 8 percent (Medicare Payment Advisory Commission, 2001a), (Medicare Payment Advisory Commission, 2001b).

All the reductions, however, cannot be attributed to the changes in the BBA.

Approximately concurrently with the BBA the federal government greatly increased the resources it devoted to anti-fraud and anti-abuse efforts in the Medicare program.64 These resources were especially targeted on home health services, but they were applied throughout the program. One sign of the anti-fraud efforts was that in 1998 the Case Mix Index for inpatient admissions fell for the first time since the PPS began in 1984. Later analysis by HCFA showed that hospitals were coding similar cases in lower-weighted DRGs (a “downcoding” of 0.5 percent), the first time this had been observed.65

Because of the confounding of the BBA reimbursement changes and the anti-

fraud efforts, a good estimate of the savings attributable purely to the BBA cannot be made. But the low Medicare outlays in 1998 and 1999 were well below the outlays projected at the time of the BBA, leading providers to lobby for higher payments on the grounds that the BBA was unexpectedly harsh.66 Partly because the long-run fiscal outlook for the Part A Trust Fund had greatly improved (see below), the Congress in 1999 in the BBRA and again in 2000 in the BIPA increased payment rates for almost all providers above what the BBA had put in statute. In particular, it increased payments to SNFs in 1999 and again in 2000, and it also increased monies for home health in 2000. It also stayed the 15 percent reduction in home health rates that the BBA called for when the home health prospective payment system was implemented.

Nonetheless, the givebacks in 1999 and 2000 were modest on the scale of the

entire program. The Congressional Budget Office estimated that BBRA givebacks would cost $17.2 billion over 10 years; the BIPA givebacks were estimated to cost substantially more, $81.5 billion over 10 years (2000 was an election year).67 Nonetheless, the givebacks were small compared with Medicare spending and also with the BBA cuts. In January 2000 the CBO expected total Medicare spending over the next ten years to be $3,226 billion; using that as a denominator, the givebacks

64 The effort was known as Operation Restore Trust. 65 MedPAC, like its predecessor commission ProPAC, explicitly considers the degree of coding change in its update recommendations. The intent is to pay for true change in case mix but not for coding the same patient in a higher weighted DRG. Following methodology developed by (Carter, et al., 1990), HCFA assessed coding change by having expert coders, who were blinded to the year of the case, simultaneously code a random sample of cases from two different years. The difference in the Case Mix Index for these two years is used to measure the true case mix change. 66 In August 1997 the Congressional Budget Office estimated the BBA would save $112 billion between 1998 and 2002. By July 1999 their estimated savings from the BBA and from the increased anti-fraud efforts for the same period were almost twice as large, $217 billion. See http://www.cbo.gov/showdoc.cfm?index=1553&sequence=0&from=5. 67 See http://www.cbo.gov/showdoc.cfm?index=1681&sequence=0&from=5 and http://www.cbo.gov/showdoc.cfm?index=2799&sequence=0&from=5.

28

raised Medicare spending above what it otherwise would have been by about 3 percent. The givebacks were also small in comparison with the BBA cuts; in August of 1997 the CBO expected the BBA to save $385 billion over the following ten years. Two years later it had doubled its estimate of savings in the 1998-2002 period from both the BBA and the anti-fraud efforts.68

In sum, the BBA almost certainly reduced Medicare outlays, which helped restore

the long-run fiscal health of the program (see below). It increased the incentives for efficient production by mandating the development of prospective reimbursement systems for post-acute care and for hospital outpatient departments, thereby ending cost-based reimbursement virtually throughout the Medicare program. It addressed the geographic differences in payment in the Medicare+Choice program, though not in traditional Medicare, and it changed physician payment from a course that appeared unsustainable. In my judgment, however, the changes the BBA made have left us with a new set of serious problems. Before turning to issues around the long-range financing of the program, I discuss the problems the BBA created in three areas, post-acute care and hospital outpatient reimbursement, the Medicare+Choice program, and physician payment.

Some Current Issues Post-Acute Care and Hospital Outpatient Departments As noted above, HCFA has implemented new prospective schemes for home

health agencies, SNFs, rehabilitation facilities, and outpatient departments, and did so on or near the very tight deadlines specified in the BBA. Nonetheless, it is not clear that the new schemes will function tolerably well. The problems are fourfold.

First, and fundamentally, for all the post-acute services the new payment methods

largely do not pay more for additional services; in contract theory jargon they are reasonably high powered. Our ability to specify a priori and monitor a desired bundle of services is markedly less for these services, however, than for inpatient services. Economic theory suggests that payment should be lower powered, the less well we can specify a priori the bundle of services and monitor that it was delivered, so these changes seem to have gone in the wrong direction. Put another way, providers now have a financial incentive to underserve or stint on these services. The home health reimbursement system is particularly high-powered, with zero marginal revenue for all visits past five within a 60-day period.69 And monitoring is particularly difficult in home health, where auditing what actually happened during the visits is a daunting task.

68 See the sources two notes above, as well as http://www.cbo.gov/showdoc.cfm?index=1820&sequence=0&from=1, Table 4-6 and http://www.cbo.gov/showdoc.cfm?index=308&sequence=0&from=5. 69 There is a modest outlier provision.

29

Second, the new payment systems set different prices for the same service in different settings, yet many patients can obtain the service in multiple settings. Physical therapy, for example, can be given in a hospital outpatient department, in a SNF, in a rehabilitation hospital or unit of an acute care hospital, or if the patient is well enough to go home, through home health.70 Both the level and basis of reimbursement, however, vary considerably across these sites of service. Because the most intense treatment is given in the rehabilitation hospitals and units and the least intense at home and because the new systems were implemented in a budget neutral fashion for each site, reimbursement is highest in the rehabilitation facilities and lowest at home. Furthermore, as described above, patients in rehabilitation facilities are reimbursed per case (by the stay), SNF patients are reimbursed per day, and home health patients per 60-day episode. As a result, the payment system is far from neutral as to where a given patient should be treated. We have little data, however, on the degree to which these non-neutral incentives have affected where and how patients are cared for.71

Third, 18 percent of patients discharged from a hospital use multiple post-acute

providers (Medicare Payment Advisory Commission, 1999b). In effect, many, if not most of these patients receive part of their treatment at one site (e.g., a SNF) and then transfer to another site to complete their treatment (e.g., home). Accounting for partial episodes of treatment in various sites introduces substantial additional complexity, as well as possibilities for gaming.

Fourth, the difficulties of securing reliable data on what services were delivered at

the patient level have hampered the development of case-level adjustments. Partly for this reason MedPAC’s judgment in March 2001 was that the RUG system, which was initially developed for chronic long-term care, could not be adapted for the post-acute care of SNF patients. It therefore recommended that HCFA stop work on refining the RUG system and focus on developing a new system (Medicare Payment Advisory Commission, 2001a). Although this view is controversial, it indicates the depth of the problems with developing new post-acute payment methods. There has been less experience with the home health HHRGs, since they have been in place less than a year, but with the large reduction in the number of home health visits after the BBA, there is no reason to think the weights based on historical data are appropriate.72

70 For the sake of completeness, it can also be given in a Comprehensive Outpatient Rehabilitation Facility (CORF) and an Outpatient Rehabilitation Facility (ORF). 71 The ability to move patients into rehabilitation hospitals and units, where the payment is highest, is limited by regulations. 75 percent of patients in rehabilitation hospitals and units have to be in one of ten DRGs, and all patients have to receive an average of three hours of active therapy per day. Not all patients, especially the frail elderly, can tolerate three hours of daily therapy. 72 The new adjustment system for rehabilitation hospitals and units, FIM/FRG, has been developed and used by those hospitals for several years for case-mix adjustment for quality purposes; its prospects therefore seem better to me than those of RUGs or HHRGs, although the problem of reconciling per case rehabilitation payments with the per diem and per episode payments of the other systems remains. In the short run the regulation requiring three hours of therapy per day described in the previous footnote is likely to be important in limiting undesired substitution toward rehabilitation sites.

30

In my view the ultimate solution for post-acute services will be to bundle the

payment for them with the DRG payment to the hospital, with a reasonably high fraction of cases receiving some kind of payment at the margin (i.e., lowering the power of the system). Freestanding post-acute providers, however, are fiercely opposed to such a solution because the payment would probably go to the hospital and they would then become contractors to the hospital. Although they have carried the day for now, I doubt very much that any other system will work very well.73

The problem of payment rates for the same service that vary by site is not limited

to post-acute care; it also appears in reimbursement for ambulatory services. Virtually all of these services can be delivered in at least two of three settings: the hospital outpatient department; the Ambulatory Surgery Center, or the physician’s office. But the rates Medicare reimburses differ substantially among these sites (Table 11). Additionally, some patients can be treated on either an inpatient or outpatient basis. Thus, the payment system is far from neutral in this domain as well.74 Moreover, the rules for determining whether the building adjacent to the hospital will be reimbursed as the outpatient department or the medical office building, or for that matter the Ambulatory Surgery Center, are ambiguous. As with post-acute care, the importance of this non-neutrality of payment across sites is unclear.

Medicare+Choice There were three salient effects of the BBA on the Medicare+Choice program:

reductions in reimbursement in response to selection; an effort to narrow geographic differences in spending within the Medicare+Choice program; and the introduction of diagnostic-based risk adjustment. I deal with these three effects in turn.

Taking back the profits of selection. In response to the reductions in payment to

HMOs mandated by the BBA, HMOs have pulled out of several counties on January 1 of each subsequent year (Table 12).75 These pullouts attracted a great deal of

73 If hospitals were given a larger DRG payment along with responsibility for post-acute care and discharged this responsibility by contracting with certain providers, there could be a court challenge that such contracting abridged patients’ freedom of choice of provider. At issue would be whether the patient can choose any post-acute provider or whether, when using a hospital, the patient agrees to use certain providers, as for example the patient does with respect to the laboratory with which the hospital contracts for services. 74 One argument for non-equal payments is that patients in the outpatient department may be sicker; sicker patients may be treated there because hospital backup services are available. Nonetheless, it seems unlikely that this feature could account for anything like the differences in reimbursement shown in Table 11. 75 Perhaps surprisingly, HMO pullouts were disproportionately in areas that received the largest payment increases (Medicare Payment Advisory Commission, 2001a). This reflects HCFAs changing its interpretation of one part of the law after the BBA. Prior to the BBA HCFA had allowed HMOs to offer different supplemental benefits in each county, partly reflecting the different reimbursement rates in each county. After the BBA HCFA believed it no longer had the authority to do so and required uniform benefits throughout a service area, which was usually several counties. Particularly when a service area consisted of a city and some surrounding rural counties, HMOs sometimes opted to withdraw from the rural

31

publicity for at least two reasons. Some saw competition among health plans as the future of the program, but the withdrawals cast doubt on whether this was realistic. Second, growth in enrollment before the BBA had been rapid and no plans had previously pulled out. As a result, most thought growth would continue. The Congressional Budget Office (CBO), for example, had projected in 1997 that HMO penetration would reach 30 percent by 2005, a projection that four years later looks distinctly optimistic (Congressional Budget Office, 1997).

The pullouts had serious consequences for some beneficiaries. Those enrolled in

the HMOs that no longer contracted with Medicare potentially had to change physicians and probably lost some supplementary benefits. On the other hand, it would be poor policy for the program to pay a rate sufficiently high to keep every health plan in business, no matter how poorly run. And enrollments in HMOs tended to climb back in the months after the January 1 pullouts, so that although the share of beneficiaries enrolled in Medicare+Choice has fallen from its high of 16 percent, but remains at around 15 percent (Figure 2). Nonetheless, enrollment is now stagnant after growing rapidly in the 1990s.

Payment floors. The Congress established floors on payment to health plans in

response to perceived geographic inequity; in particular, beneficiaries in the high rate areas were getting drug benefits and those in low rate areas were not or were getting fewer benefits. Perhaps some conservatives may have also had an eye toward reducing pressure for enacting an outpatient prescription drug benefit by giving beneficiaries an option to join HMOs to obtain drug coverage (see below). But the effort to address geographic equity has the potential to change the nature of the Medicare+Choice program radically because it has unbalanced local health care markets. That is, because the Congress did nothing about variation in spending in traditional Medicare, it made traditional Medicare more attractive in the high rate areas and less attractive in the floor areas.

In the high rate areas rate increases for Medicare+Choice plans are now limited to

3 percent per year (up from 2 percent in the BBA). To the degree costs rise faster than this and the plan market is competitive, which for the most part it is in the high rate areas, supplementary benefits will be taken off the table, and beneficiaries will tend to drift back toward traditional Medicare.76

The Congress put the floors in place in an effort to attract HMOs – preferably

bearing drug benefits – to small population counties that lacked them. This policy assumed that the reason HMOs were not in these areas was the low rate of payment, but this assumption is likely wrong because the structure of small markets is not attractive to HMOs. In many small areas there is one local hospital, and there may be

counties, where reimbursement was considerably lower, rather than provide the same supplementary benefits that were being provided to urban residents. And rural counties, as described in the text, received the largest updates. 76 This drift will be mitigated to the degree that those opting to go back to traditional Medicare cannot obtain Medigap without medical underwriting.

32

only one type of a given specialist. Because the threat to shift business to another provider is not credible in many such instances, HMOs are unable to obtain rates below the Medicare fee schedule and thus have difficulty making a profit. Similarly, a threat to terminate a provider from the HMOs network for not following guidelines is not credible. Consistent with this argument, HMOs have not entered in response to the higher floors.

But in the floor counties there is a good deal more money now in

Medicare+Choice plans than in traditional Medicare, upwards of $2000 per beneficiary per year in some counties. These additional funds make the private fee-for-service option attractive in the floor counties. There were very few utilization controls within traditional Medicare, so equivalent medical care can be delivered under the private fee-for-service option at considerably less than the payment. In short, there are rents to be appropriated. Economics would predict that providers in those areas that have market power will, in the end, capture most of these rents, although until there is competition among private fee-for-service insurers, many of the rents will accrue to the early insurer entrants in this market. Some of the rents may go to beneficiaries to induce them to leave traditional Medicare, but physicians in the floor counties may simply stop participating in traditional Medicare and tell their patients to choose the private fee-for-service plan. In non-floor counties, private fee-for-service is unlikely to succeed. Because almost all physicians accept Medicare beneficiaries in those counties, there is no incentive for beneficiaries to pay more for a private fee-for-service plan. In the floor counties, however, beneficiaries will not have to pay more.

Thus far only one private fee-for-service plan has entered the market, although it

has entered in 35 states. As theory would predict, it has entered disproportionately in floor counties. HCFA is currently reviewing the application of a second private fee-for-service plan.

Risk Adjustment. HCFA did carry out the mandate of the BBA by proposing a

risk adjustment method based on Diagnostic Cost Groups (DCGs). Like the DRGs for hospital patients, DCGs group patients by their diagnosis (if any), and pay health plans more for those enrollees with diagnoses that are costly to treat. Confirming the existence of favorable selection, HCFA estimated that the average health plan would lose 7 percent of its revenues when this method was fully implemented. Health plans, arguing that reimbursement was already inadequate and forcing some of them to leave Medicare (i.e., citing the pullouts), lobbied for budget-neutral risk adjustment. This they did not obtain. But the health plans also argued correctly that the risk adjustment method HCFA proposed gave them an incentive to distort their treatment choices. This problem arose because HCFA was forced to employ diagnostic information from the inpatient setting only.

Reliable diagnostic information is available on inpatients because the DRG and

hence hospital reimbursement is a function of diagnosis. If hospitals do not report diagnosis accurately, they are liable for criminal penalties. Although physicians are

33

to report diagnosis on their Part B claims, their reimbursement does not depend on diagnosis; as a result, there is substantial undercoding of diagnosis for office visits. Table 13 shows the probability that a Part B claim in 1995 had a given diagnosis, conditional on a Part B claim showing that diagnosis in 1994 (and that the patient was alive throughout 1995). Because of the seriousness of these diagnoses and because they are chronic diseases, almost all patients with these diagnoses should have had a physician visit in 1995 that recorded the diagnosis, but only around half show a claim with such a diagnosis.

These results posed two problems for HCFA. First, they meant the weights for a

given DCG that HCFA was using were likely in error because patients were misclassified. Second, they meant there was an enormous potential for upcoding if HCFA were to pay on the basis of diagnoses made in the outpatient setting. That is, given the new financial incentive to code accurately, many more patients would appear in higher weighted DCGs than was the case in the historical data being used to set rates on a budget neutral basis. Given that likelihood, HCFA opted to introduce risk adjustment using inpatient data only. But this gave health plans an incentive to hospitalize patients merely to record the diagnosis and hence obtain the higher payment. HCFA recognized the problem; it therefore proposed not only a four-year transition to a fully risk adjusted payment but also basing only 10 percent of the reimbursement on the risk adjusted payment in the first year. This percentage would rise to 35 in the second year, 80 percent in the third year, and in the fourth year outpatient data would be incorporated. The transition also mitigated the payment reduction from risk adjustment.

The plans lobbied to defer diagnosis-based risk adjustment indefinitely. Although

the Congress did not agree to this request, it did slow down the transition. Payment is still only 10 percent risk adjusted, so the bulk of the incentives for plans to select healthy beneficiaries within age-sex classes are still in place. Moreover, if and when outpatient data are used, there will almost certainly have to be another transition to estimate and allow for effects of upcoding.

More fundamentally, although the DCGs are a substantial improvement over the

prior methods, or at least will be once outpatient data are incorporated, no one knows whether they will be sufficient to render selection behavior negligible. One way to gain some insight into how selection incentives would change is to quantify how profitability would change given certain selection strategies. If a plan enrolled a random sample from the 20 percent of traditional Medicare enrollees who spent the least in 1991, the plan would make a profit of $2,134 per enrollee in 1992 if the risk adjustment method used only the demographic adjusters of the AAPCC (and if the plan treated the beneficiary the same as he or she was in fact treated in traditional Medicare). With the DCG method (but using diagnoses from the outpatient setting) the profit would be $424 per enrollee, still a healthy profit rate on mean spending of $3,800, but clearly much less than $2,134. At the other extreme, if the plan enrolled a random sample of the 20 percent who spent the most in 1991, it would lose an average of $4,425 in 1992 with demographic adjustment but only $1,311 with the

34

DCG method (Ellis, et al., 1996).77 We do not know the costs of selection, but it would seem that a plan that can devise inexpensive methods to select still stands to profit handsomely at the expense of the government. My judgment is that risk adjustment is still highly imperfect, and it is likely to be many years before it will be substantially better.78

For that reason I and others have proposed what is variously termed partial

capitation or supply side cost sharing (Ellis and McGuire, 1986), (Ellis and McGuire, 1993), (Newhouse, 1986), (Newhouse, et al., 1989), (Newhouse, 1996), (Newhouse, et al., 1997), (Newhouse, 1998). In its simplest formulation reimbursement to the plan would be a weighted average of what would otherwise be paid under a risk-adjusted capitation and what would be paid under traditional Medicare, but non-linear formulations are also possible. Such a formulation sacrifices some incentives for efficiency in production in order to reduce both incentives for selection and incentives to stint or underserve.79 MedPAC and its two predecessor commissions have all recommended this, but the proposal has gotten nowhere legislatively because health plans oppose it. Their reasoning is not far to seek; just as with risk adjustment, partial capitation reduces the profit from the favorable selection that health plans now enjoy.

A further issue is whether some portion of a plan’s payment – or for that matter a

provider’s payment in traditional Medicare – should be based on observed quality. There are at least two potential problems with doing so. One is the cost of collecting and verifying the information to adjust for variation in the disease mix of patients across plans and providers. A second is the multi-tasking problem, the incentive of providers to shift resources toward measured areas (Holmstrom and Milgrom, 1991).

77 A more customary, but less intuitive way to measure the gains from risk adjustment is as follows. At least 20 to 25 percent of variance in spending across individuals is predictable. This figure comes from a fixed effects model with an AR(1) term. Fixed effects account for about 15 to 20 percent of the variance, and the AR(1) term accounts for about another 4 percent. This is a lower bound on explainable variance because it does not account for time varying effects that are observable (e.g., the person has just been diagnosed with cancer). The demographic adjusters in the AAPCC only explain about 1 percent of the variance, so they are better than nothing but not by much. Inpatient DCGs, when fully implemented, explain about 6 percent of the variance; incorporating diagnoses from the outpatient setting brings this figure up to the 9-11 percent range. See (Newhouse, et al., 1989) and (van de Ven and Ellis, 2000) for more details. 78 (Glazer and McGuire, 2000) advocate over- and underpaying on observed variables such as age that correlate imperfectly with risk rather than paying expected costs at each age. I believe there are two types of problems if policy were to go in this direction. First, their model makes assumptions that may not hold in reality (i.e., profit-maximizing plans or capitated medical groups allocate treatment to members to maximize patient welfare rather than profit; plans or groups cannot configure themselves to discriminate on those variables for which payment is amplified or attenuated; for example, if children are paid less than expected costs, plans cannot or would not have few pediatricians in their networks). Second, the amount of the optimal over- and underpayment depends on the unobservable correlation of the measured variable (e.g., age) with true risk. Hence, this plan does not appear implementable. 79 The reduced efficiency in production comes in part from less reward to effort at cost reduction (Laffont and Tirole, 1993) and in part from a distortion in price between covered services and non-covered services for which there would be no reimbursement under traditional Medicare. One could in principle set up a reimbursement method for these services, but this would be administratively difficult.

35

Some private health plans, however, do base compensation of primary care physicians on patient-reported satisfaction with the physician.

Physician Payment The BBA, as mentioned above, established a formula for spending on physician

services, the Sustainable Growth Rate, that was a function of, among other things, real GDP growth. This method of setting total spending differs from the method used for other providers, where the Congress sets rates or prices with no explicit or formulaic account taken of past changes in the quantity of services or of GDP growth.

As described above, this method for controlling spending on physician services

stemmed from the rapid increase in physician spending in the 1970s and 1980s, and the view that if fees were reduced, physicians would simply order more services to offset the loss in income. Hence, the Congress believed a cap on total spending was necessary.

There was empirical support for the view that physicians increased services when

fees fell, strong enough support in fact so that when the new fee schedule was introduced in 1992, HCFA actuaries allowed for a “behavioral offset.” Specifically, the actuaries assumed physicians would offset the fee cut by increasing the quantity of those services whose fees were being reduced by 30 to 50 percent and so applied an additional reduction to the conversion factor to reach the desired spending cut (Physician Payment Review Commission, 1993).80 In fact, the actuaries overestimated. Although physicians did on balance increase those services whose fees had been reduced, they also decreased services whose price had increased. These effects approximately netted out (Zuckerman, et al., 1998).81

Nonetheless, the view has persisted that a physician can simply order services to

reach his or her desired income, so that policy must cap total spending rather than simply set prices independently of volume. I think the evidence against this view is compelling.82 But even if this view is correct, a cap is problematic as policy because of the possibility of substitution of care among sites. As shown in Table 11, hospitals

80 Later work pointed out that there was no necessary reason why physicians would limit their volume increases to Medicare patients or to those services whose fees were affected (Yip, 1998), (McGuire, 2000). 81 The view that physicians could circumvent an effort to reduce spending through fee reductions by ordering more units of service stems from the early health economics literature, which sometimes postulated a target income model for physicians (e.g., (Evans, 1974)). Such a model, of course, left open the question of how the target was established and why a maximizing physician would limit income to the target amount. After several years of what resembled theological warfare in the literature, the empirical findings of an inverse relationship between fees and the supply of services were reconciled with a maximizing model by postulating a model in which a physician had some ability to induce demand and would increase the amount of inducement if an income effect from a fee cut were large enough (McGuire and Pauly, 1991), (McGuire, 2000). Otherwise, a standard substitution effect would dominate and units of service would fall. 82 One piece of evidence is that physicians locate as standard location theory would predict, and as their numbers have increased, have located in smaller and smaller cities and towns. See (Newhouse, et al., 1982). If they could fully dial services up and down, they could continue to locate in desirable places.

36

are paid substantial amounts to cover their costs for services performed in their outpatient departments (e.g., nurses salaries, supplies). Because physicians do not bear these costs in the outpatient department but do bear them in their offices, the site-of-service adjustment described above reduces the practice cost component of physician reimbursement by 50 percent for services performed in the outpatient department and in Ambulatory Surgery Centers (relative to the physician’s office).83 But outpatient physician services move across various outpatient sites in response to technological change – and some also move from being inpatient procedures to being outpatient procedures – in ways that a fixed target cannot accommodate. MedPAC has recently recommended that the Congress abandon its approach to controlling spending on physician services and set physician fees in a fashion similar to rates for hospitals and other institutional providers (Medicare Payment Advisory Commission, 2001a). Whether the Congress will act on this recommendation is unclear.

Financing Medicare in the Long Term The Bipartisan Commission. The BBA set up a Bipartisan Commission on the

Future of Medicare, whose chair was Senator John Breaux (D-LA) and whose administrative chair was Congressman William Thomas (R-CA). The Commission’s main agenda was to consider how to finance Medicare over the longer term. Medicare had for many years prior to the BBA grown at rates far in excess of the growth rate of federal tax revenues, and in 1997 this seemed likely to continue indefinitely (Figure 4). The 1996 Report of the Trustees of the Medicare Trust Fund, the backdrop for the BBA, projected that the Part A Trust Fund would have a zero balance in 2001; the 1997 Report pushed that date back to 2005. And, if Medicare was going to be gasping for funds in 2005, matters were going to become much grimmer after the baby boomers started to swell the ranks of the beneficiaries in 2010 – and grimmer still when they started to turn 75 in 2020. In short, at the time of the BBA the long-run financing of Medicare appeared to be a very serious substantive and political problem.

Senator Breaux and Congressman Thomas favored shifting Medicare toward a

defined contribution approach that would include traditional Medicare. That is, the government would pay a lump sum, as those employers that offer multiple plans often do, and the beneficiary would be responsible for paying the marginal dollar. Such an approach has been endorsed by a wide variety of economists of varying political persuasions, including myself (e.g., (Cutler, 1995), (Aaron and Reischauer, 1995), (Butler and Moffit, 1995), (Wilensky and Newhouse, 1999)).

Those advocating this approach see several advantages. First, it would potentially

make the government neutral among choice of health plan. Under the present arrangement a lower cost plan has a limited ability to pass on lower costs in the form

83 One might ask why the reduction is not 100 percent. Physicians argued that they incurred fixed costs to maintain an office even when they were physically elsewhere, and they needed some monies to pay for these fixed costs. The Congress agreed in principle and so reduced the practice cost component of reimbursement only by 50 percent.

37

of lower premiums. Specifically, the plan may not offer rebates, so that the most money a beneficiary can save by joining a health plan is any Medigap premium plus the expected value of any remaining cost sharing.84 Any additional savings cannot come in cash but must be taken in the form of additional covered services, which the beneficiary may or may not value at their cost to the plan.85 Second, a defined contribution approach may lead to more efficient production by freeing up pricing underneath the level of the plan rather than relying on the administered pricing systems of traditional Medicare and the types of distortions described above.86 Freeing up pricing underneath the plan may also avoid some the problems described below in introducing new products into Medicare’s administered price schemes.

Those opposing a defined contribution approach worried that traditional Medicare

would become more expensive under this proposal because better risks would tend to leave it. In the extreme, traditional Medicare could go into a death spiral. In other words, true neutrality among competing health plans assumes adequate risk adjustment, something that at present requires a leap of faith – maybe several leaps of faith – although use of partial capitation can reduce the load that risk adjustment needs to bear. Additionally, there were concerns about geographic adjustment; if enrollees from low cost areas were pooled with those from high cost areas with no geographic adjustment, for example, they would be worse off. Finally, there were concerns among some that low-income Medicare beneficiaries not eligible for Medicaid could be coerced into plans with skimpy benefits or high cost sharing.

The Commission of 17 members operated under rules that required a

supermajority of 11 to make formal recommendations. Initially it was hoped there might be a deal that involved changing Medicare to a defined contribution approach and adding an outpatient prescription drug benefit. But only ten votes could be mustered for the chairs’ proposal that embodied defined contribution principles and addressed the drug issue. Specifically, the chairs proposed a new fee-for-service option involving private sector insurers partnering with HCFA to offer policies with a stop-loss provision and drug benefits, although there would be no commingling of money or management between HCFA and the private companies. Medicare HMOs would simply have the actuarial value of the stop-loss and drug benefits added to the AAPCC, so that they could offer additional benefits (e.g., a higher drug maximum).

84 Those with retiree health insurance, 36 percent of beneficiaries (Table 1), are unlikely to find anything other than traditional Medicare attractive unless they can receive monetary savings from another choice. For them to receive such savings most employers would have to modify their retiree plans to allow some kind of cash out, but under present arrangements, employees appear to have little interest in such a cash out. Defined contribution arrangements might stimulate more demand from employees to make such a change, although this is highly speculative. The Benefits Protection and Improvement Act of 2000 includes a provision to allow premium rebates starting in 2003. 85 This statement needs qualification to the degree the plan can obtain a lower price than the individual for certain services. 86 In addition to the distortions that arise even in a benevolently administered scheme, political distortions arise. It is difficult, for example, to exclude any provider from traditional Medicare, even a poorly performing one, because, among other reasons, doing so might reduce the flow of federal dollars to the provider’s community and Congressional district. In the case of rural communities, the hospital can be the largest employer in the town, and if it closed, the community itself might have a difficult time surviving.

38

When 11 votes for this proposal could not be found, the Commission did not file a formal final report, but the Chairmen’s Report, transcripts, and other Commission documents can be found on the Commission’s website.87

At the time the Commission disbanded in 1998, the Administration attacked the

suggestion that Medicare move to a defined contribution framework. This may have resulted from pressure from Congressional Democrats who wished to run on this issue, hewing to the successful 1996 strategy of not changing traditional Medicare for the 85 percent of the beneficiaries who had elected it. In 1999, however, the Administration did something of an about face and introduced a proposal that could probably have been compromised with the proposal that attracted 10 votes from the Commission.

The Administration’s 1999 proposal made one important change from the

proposals that the ten-person majority on the Bipartisan Commission favored. Whereas the Commission had proposed that the government’s contribution be increased in the future at the rate of the weighted average premium across plans, the Administration proposed that it increase it at the rate of increase in the premium for traditional Medicare. Thus, the Administration protected beneficiaries who chose to remain in traditional Medicare against an increase in their Part B premium. Initially, however, both proposals kept the Part B premium at the level of current law. If there were a total premium for Parts A and B together, current law implied an 88-12 percent split between the government’s contribution and the beneficiary’s share.88 This left open a possible compromise of making the 88-12 split less favorable to beneficiaries over time, which seems likely to happen if health care costs increase at historical rates (Fuchs, 2001). Second, if enrollees chose a lower cost plan, they would only receive 75 percent of the savings, rather than the 100 percent under the Commission’s proposal, though under both plans if they chose a more expensive plan they would pay 100 percent of the excess.

The Administration’s proposal, while an important departure for a Democratic

Administration, came a little over a year before the 2000 elections, a time when Congressional Democrats – and probably a number of Republicans – did not want to take up major reform of the Medicare program. As a result, nothing came of it.

Moreover, the increase in payroll taxes from the economic boom along with the

unexpectedly large reductions in spending from the BBA and the anti-fraud efforts meant the long-term finances of the program looked dramatically better. By 1999 the Trustees Report projected that the Part A Trust Fund would not go to a zero balance until 2015, the 2000 Report pushed that date out to 2025, and the 2001 Report set a date of 2029 (Board of Trustees, 2001a). As the date receded, the political impetus for large-scale reform decreased. Changing Medicare to a defined contribution approach involved sufficient political pain that many members of Congress were happy to leave this job to their successors. Although President Bush campaigned on

87 See http://medicare.commission.gov/. 88 The beneficiary’s share under current law was the Part B premium.

39

the issue of converting Medicare to a defined contribution plan, I believe its chances of enactment in this Congress are slim unless a deal is struck to add a prescription drug benefit in exchange for moving Medicare to a defined contribution arrangement.

I should add that in my view the Trustees are overly optimistic about the long-run

rate of Medicare cost increase and that therefore the long-run financing problem is more serious than their estimates imply. The projections using the intermediate assumptions, which are those commonly cited and the basis for the dates cited above, assume that between now and 2025 real hospital payments per beneficiary will increase annually at approximately the rate of per capita GDP plus 1 percentage point, whereas the rate of increase between 1975 and 1996 was GDP plus 2.25 percentage points (Board of Trustees, 2001a).89 Even the Trustees’ “high” assumption assumes that costs only grow about 2 percentage points more than GDP. The Trustees do not provide an estimate of when the Part A trust fund will have a zero balance under their high assumptions, but they do say that under those assumptions, costs grow faster than income by 0.45 percent annually between 2001 and 2025, whereas income grows 0.04 percent faster than cost on their intermediate assumptions.

In the case of Part B spending, the Trustees seem even more optimistic. They

assume annual spending per beneficiary increases only 0.7 percentage points faster than the rate of GDP between now and 2025. Between 1975 and 1996, however, spending per beneficiary on Part B grew a full 4.15 percentage points more than GDP. Although it is highly unlikely that Part B spending per beneficiary will grow as rapidly in the future as in the past, the Trustees’ projection seems improbable to me. Even the Trustees’ “high” assumptions for Part B envision a rate of growth that is only two percentage points above GDP (Board of Trustees, 2001b).

Finally, none of the Trustees’ projections includes a prescription drug benefit.

Depending on the structure of the drug benefit, it could add perhaps 10 to 20 percent to Medicare spending on a once-and-for-all basis. Moreover, most analysts expect the rate of increase in drug spending to exceed that on other health care services, so the steady-state rate of growth in Medicare would also probably increase if a drug benefit were enacted.

In sum, although no one can know what will happen in the future, I find the

Trustees’ “high” projections a more plausible scenario than their intermediate projections. Thus, I expect Medicare’s long-run financing problem is a serious one indeed.

Paying for Technological Change. Two quite different issues are created by

welfare-increasing but costly technological change, something that seems to happen almost daily in medical care. The first, likely to be on the agenda of most future Congresses, is how the burden of paying for such change will be shared between the

89 I use 1996 as an endpoint to avoid correcting for the BBA mandated shift of home health spending from Part A to Part B that was described above, and 1975 as a beginning point to be past the inclusion of the disabled and those with End Stage Renal Disease in the program.

40

elderly and the non-elderly. The second is how Medicare’s administered price methods reimburse for new products, a traditional problem for administered price systems.

Who Pays? Under current law most of the cost of technological change inevitably

falls on the non-elderly, since the program is mainly tax financed.90 In a defined contribution approach, the issue of the division between the elderly and non-elderly turns on how the government’s contribution will be updated to account for cost-increasing change.

Two formulaic proposals for updating a defined contribution both appear

unsatisfactory. One would index the government’s contribution by a medical care price index. Setting aside the important upward biases in the current official indices ((Berndt, et al., 2000), (Newhouse, 2001)), conceptually such a proposal would put the entire burden of costly new products on the elderly. A second approach would index the contribution by a measure of change in private insurance premiums, which would put almost all the burden on the non-elderly, since they pay a disproportionate share of the taxes. In practice, at least with a defined contribution approach, the outcome is likely to be somewhat in between these two approaches, although a proper price index might be a reasonable lower bound for an update.

Reimbursement. Technological change also poses a difficult problem for

Medicare’s administered price systems, especially the physician and outpatient department systems, which are very disaggregated relative to the inpatient PPS. Simply put, if there is no billing code in those systems for a new product, there is no reimbursement. The situation is only marginally better in the hospital system, where there is no immediate change in the DRG payment if a cost-increasing but welfare raising product comes to the market. In the hospital case, however, if the product is sufficiently better that it is introduced despite no incremental reimbursement, its costs will begin to be reflected in the updates to the DRG weights (relative prices).91

In the case of the outpatient reimbursement systems, obtaining a billing code may

require a coverage decision, but even if it does not, there is generally a substantial lag. The BBRA attempted to rectify the resulting bias against new products by mandating pass through payments for certain drugs and devices in the new hospital outpatient payment system, implemented in August 2000. But allowing the hospital to simply pass through its costs for specific products invites manufacturers, especially of drugs and devices with high Medicare outpatient shares, to set high prices and undercuts price competition among substitute products. To limit potential federal spending under this provision, the Congress capped these payments at 2.5 percent beginning in 2003 and 2 percent thereafter, but political pressure precluded imposing any cap before that time. It seems likely that spending in this area could rise substantially;

90 There is a bit of flexibility in dividing the burden in setting the Part B premium. 91 This happens because the annual updates in the weights reflect relative charges per case across different DRGs. See (Kane and Manoukian, 1989) for a case study of a product whose diffusion the hospital PPS deterred, at least in the short run.

41

indeed, whether the caps now in law will in fact be imposed at these levels seems problematic. For further discussion on this point, see (Medicare Payment Advisory Commission, 2001a).

A Prescription Drug Benefit The two principal services Medicare does not cover are outpatient prescription

drugs and chronic long-term care.92 Proposals to cover long-term care within Medicare have occasionally been made, but the cost, as well as the view that Medicaid is available as coverage of last resort, has deterred serious policy consideration. Cutler and Gruber in their chapter in this volume discuss the Clinton administration initiatives in long-term care.

By contrast, adding a prescription drug benefit to Medicare has not only been

considered for many years but was in fact enacted as part of the Medicare Catastrophic Coverage Act of 1988.93 Starting in 1991, the Act would have provided a benefit of 50 percent coinsurance above a $600 deductible; in 1992 the deductible would have risen to $652 and the coinsurance would have dropped to 40 percent. In 1993 the coinsurance was to fall to 20 percent. The intent was to then raise the deductible so that about a sixth of the beneficiaries would exceed it in any one year.

The Act was to be financed entirely by the elderly, through an increase in the Part

B premium and a surcharge on beneficiaries with incomes greater than $40,000. The surcharge was as much as $800 in 1989 and was projected to rise to over $1,000 in 1993. Many higher income beneficiaries already had drug coverage through employer-provided retiree health insurance, and were unenthusiastic about paying the surcharge for no additional benefit to themselves. Their political opposition was strong enough so that the Congress repealed most of the provisions of the Act a year later, including the drug coverage. Thus, the drug benefit never went into effect.

The 1993 Health Security Act proposed a more generous Medicare drug benefit

than the earlier Catastrophic Act, namely a $250 deductible, 20 percent coinsurance, and a $1,000 stop-loss provision. The deductible and stop-loss amounts were to be indexed so that slightly over half the beneficiaries would receive some benefits. The benefit was to be added to Part B, with about 75 percent of the financing from general revenues and about 25 percent from additional Part B premiums. The Act also proposed mechanisms to control drug prices. Manufacturers were to give Medicare rebates equal to the greater of the difference between average wholesale and retail prices or 17 percent of retail prices, whichever was greater.94 There was an additional rebate for drugs whose price increased faster than inflation. To address the incentive this offered to price new drugs higher than they would otherwise be priced, the Secretary was authorized to exclude new drugs from coverage if an agreement on price could not be reached.

92 Prescription drugs given to inpatients are covered and are reimbursed as part of the DRG payment. 93 The material in this section draws on (Medicare Payment Advisory Commission, 2000b). 94 Generic drugs and drugs purchased by managed care plans were exempt.

42

Not surprisingly, the pharmaceutical industry intensely opposed these provisions

to control its prices, arguing that they would deter new drug development. Although the principal cause for the defeat of the Health Security Act was its employer mandate to provide insurance, the pharmaceutical provisions were a contributing factor.

The impetus for a drug benefit has remained, however, spurred by the increased

spending on drugs in recent years and by price discrimination whereby those without drug insurance pay higher prices.95 Largely because of the growth in the number of efficacious drugs, Medicare beneficiaries spent 4.1 percent of their income on prescription drugs in 1998, up from 2.4 percent in 1988 (Medicare Payment Advisory Commission, 2000b), (Berndt, 2001). The elderly disproportionately use drugs; although only 13 percent of the population, those over 65 account for more than a third of the (domestic) spending on drugs.96

Responding to the financial burden of drug spending for the elderly, as well as

potential distortions in care from failing to cover drugs, the Clinton Administration in 1999 introduced a proposal for a Medicare Part D that would cover drugs. Coverage was to be offered by private health insurers or Pharmacy Benefit Managers (PBMs). Insurers would be allowed to compete for a single contract for the business of a local area. Benefits were considerably less generous than what the Administration had proposed in the Health Security Act; beneficiaries would pay 50 percent coinsurance, up to a maximum of $2,000 initially, after which there was no coverage (i.e., a maximum of $1,000 in government payments). The $2,000 maximum was to increase to $5,000 by 2008. There was no deductible. Coverage would be voluntary, but general revenues would subsidize 50 percent of the premium to reduce the burden on the elderly and to combat selection. As a further measure to reduce selection, beneficiaries would only be allowed to purchase drug insurance when they first became eligible for Medicare or if their employer dropped drug coverage from retiree health insurance.

The concern over selection certainly seemed warranted, given the experience in

the individual Medigap market. The additional premium for those Medigap policies that cover drugs, Medigap plans H, I, and J, exceeds the value of the benefit for those who spend the maximum amount on drugs (the maximum covered amount is $1,250 or $3,000, depending on the policy). Table 14 shows a comparison between premiums for Plans C and I in five cities; the differences in premiums are much greater than the $500 benefit that someone spending the maximum on drugs would obtain, implying that Policy C is close to strictly dominating Policy I because of selection.

95 As described below, Pharmacy Benefit Managers (PBMs) induce a greater elasticity of demand for given drugs, thereby lowering price for those they insure relative to those without drug insurance; see (Frank, 2001). 96 The average person over 65 had 18.5 scripts in 1995; the average person under 65 had 7.3 (calculated from (Davis, et al., 1999), Exhibit 2, and (U.S. Bureau of the Census, 2000), pp. 13, 110).

43

To assist the low-income elderly, the Administration proposed that both premiums and the 50 percent coinsurance would be fully subsidized for those elderly with incomes below 135 percent of the federal poverty level. There would be partial premium subsidies for those with incomes between 135 and 150 percent of poverty. To prevent crowdout of existing retiree health insurance, employers who offered drug coverage to retirees that was at least as generous as the Medicare benefit would receive a subsidy equal to two-thirds of the Medicare benefit. HMOs would also be given the value of the benefit as an addition to their reimbursement, so whatever drug coverage they had in place could be improved. The Administration estimated the cost of its proposal at $118 billion over ten years; CBO estimated substantially greater costs of $168 billion. Steady-state costs would be substantially more than either estimate, because of the phase-in of the benefit through 2008.

In addition to the demands on the budget, this proposal and other proposals for

Medicare drug coverage raise five issues that deserve further discussion: 1) The lack of stop-loss coverage; 2) The universality of the benefit; 3) Possible selection among competing insurers; 4) The provisions to reduce crowdout of retiree health insurance; and 5) How much authority insurers or PBMs would be given to exclude certain drugs in order to achieve lower prices.

Stop-loss provisions and cost sharing. The front-end nature of the cost sharing in

the Administration’s proposal violated elementary insurance principles, but would have paid some benefit to the 86 percent of beneficiaries who have at least one prescription in a year rather than the sixth of beneficiaries who would have received some payment under the deductible of the Health Security Act. On the other hand, the 6 percent of beneficiaries who spent more than $3,000 on prescription drugs in 1999 would have been left with open-ended liability.97

Senator Kennedy (D-MA) and Congressman Stark (D-CA) introduced a bill that

was considerably more generous than the Administration’s proposal, as well as more in accord with traditional insurance principles. It had a deductible of $200, coinsurance of 20 percent rather than 50 percent, and a stop-loss feature, set initially at $3,000 per year. I have not found a cost estimate for this proposal, but in 1999 CBO estimated a cost of around $30 billion per year for a similar but somewhat less generous proposal.98 Subsequently the Administration proposed earmarking $35 billion from 2006 to 2010 for catastrophic drug spending, but the details of this benefit were unclear.

More generally, the cost sharing provisions in the Administration bill were much

greater than most drug coverage for the under 65. Under 65 persons with drug benefits typically pay a modest copayment if they use a generic drug or a drug that is on a formulary, for example, $5 or $10 for a month’s supply. (A formulary is a list of

97 Another 14 percent spend between $1,500 and $3,000 (Gluck, 1999). 98 The proposal that CBO costed was a policy with a $250 deductible, 20 percent coinsurance, and a $4,000 stop-loss feature. See the testimony of Dan Crippen before the Senate Finance Committee, March 18, 1999.

44

favored drugs, usually those within a therapeutic class such as anti-hypertensives for which the plan has negotiated a low price.) Off-formulary drugs may carry a copayment of $25 or even $50 for a month’s supply. The Administration cost sharing provisions, as well as subsequent proposals by others, were less generous than this to keep the budget costs down, but the disparity with the benefits among the under 65 means that there is likely to be continuing political pressure to reduce the cost sharing, should any Medicare drug benefit be enacted. Reductions in cost sharing, however, if not compensated for by increases in premiums, raise the issue of the division of the burden of Medicare between the elderly and taxpayers.

Provisions for the low-income population and the universality of the benefit. The

Administration would have required Medicaid to pay both the premiums as well as the 50 percent coinsurance for those with incomes under 135 percent of poverty, about 30 percent of the Medicare beneficiaries.99 The average federal share of the Medicaid program is around 57 percent, with the remainder falling on states; thus, the proposal would have created a substantial new burden for states to cover the drug spending of those between 100 and 135 percent of poverty.100 Whether Congress would have imposed this additional financing requirement on states is problematic.

Senator Breaux and Congressman Thomas, on the other hand, proposed limiting

the drug benefit to the low-income population and providing no benefits to those with incomes above 135 percent of poverty. This would, of course, have substantially reduced the cost of the plan and would also have largely avoided the crowdout issue with respect to retiree health insurance, because few low-income beneficiaries have such insurance. It would, however, have violated the social insurance principle of universality upon which Medicare is based. Furthermore, there is little correlation between income level and drug coverage currently, so substantial numbers of Medicare beneficiaries would have remained without drug coverage (Table 15).

Selection among competing insurers. In structuring competition among insurers,

the Administration favored competition for a local contract, with the winner achieving a temporary local monopoly, for example, for three years, analogous to the competition for being a Medicare carrier or intermediary. I agree with this approach (Huskamp, et al., 2000). Under this arrangement the government would specify classes of drugs, and at least one drug from each class would have to be covered. Such an arrangement should achieve competitive pricing while preventing selection.

99 For those between 135 and 150 percent of poverty there would have been premium subsidies. In addition to the notch from dropping coinsurance subsidies at an income of 135 percent of poverty, the premium subsidies would have implied a non-trivial addition to the marginal tax rate between 135 and 150 percent of poverty. The actuarial value of the benefit was estimated to be around $600 in 2002 and $1,100 in 2008 (McClellan, et al., 2000), and the federal poverty threshold in 2000 for a single person over 65 was $8,259 and for two adults was $10,409. Thus, the increase in the marginal tax rate would have been on the order of 40 to 50 percentage points in 2002, and considerably more by 2008, depending on the rate of increase in the poverty threshold. 100 Medicaid already covers drug costs for those below 100 percent of poverty.

45

An alternative is that insurers or PBMs compete for individual beneficiaries. Under this arrangement it is unclear whether the government would require that any competing plan cover at least one of a certain type of drug (e.g., anti-depressant, anti-arthritis drug). If there were such a requirement, prices Medicare pays to drug manufacturers should be similar to competition for a contract, but multiple plans would imply additional marketing costs for seemingly little gain. If there were not such a requirement, I believe selection will divide the risk pool; certain beneficiaries would opt for the plan that, for example, includes Viagra and excludes anti-diabetic drugs, while diabetics would need to be certain that their plan covered the drugs they needed.

The structure of competition at the retail level is also important, because

distribution costs account for about 20 percent of the cost to the consumer. Most current drug plans among the under 65 use networks of pharmacies that are chosen in part on the basis of price (i.e., the price of the drug will be substantially more at a non-network pharmacy). In determining the number of pharmacies in the network, there is usually a constraint to ensure access, such as a certain percentage of beneficiaries living within a certain distance of a network pharmacy. Use of a network of pharmacies could help Medicare minimize retail costs, but for this to achieve competitive price levels, some pharmacies would have to be excluded, a politically problematic outcome.

Provisions to reduce crowdout. Thirty-one percent of beneficiaries have drug

coverage through their former employer. The Administration sought to keep this coverage in force by offering a subsidy of two-thirds of the value of the Medicare benefit to employers that provided a benefit at least as generous as the Medicare benefit.101 The administration assumed that the remaining third of the cost would be covered through the tax deductibility of the drug benefit. Although the subsidy would have arguably kept existing insurance in force, it would not have prevented crowdout in a fiscal sense. By assumption, the full cost of the plan for those with retiree health insurance would be covered entirely by the government, two-thirds by the on-budget subsidy costs and one-third by the tax expenditure. Without such a provision, however, one would expect employers to drop or restructure their retiree health insurance if a Medicare drug benefit were enacted. Thus, it appears that crowdout is simply part of the price of a universal Medicare drug benefit.

Price determination for pharmaceuticals. Among those under 65, 70 percent of

drug coverage is contracted to PBMs, firms that specialize in drug coverage, and most past and current proposals for a Medicare benefit, including the Administration’s proposal, envision using PBMs as well. As already described, PBMs employ differential copayments to direct consumers to those drugs on their formulary. Formularies lower drug prices by increasing the elasticity of demand that manufacturers face; they are the drug analog of a network of physicians and hospitals. Because the PBM market is reasonably competitive, most of these discounts are

101 If their employer received a subsidy, beneficiaries would not have been eligible for the Medicare drug benefit.

46

passed on to consumers. I have advocated an analogous procedure for Medicare, namely the use of reference pricing; under this procedure drug manufacturers would submit bids within classes of drugs, and consumers would pay the entire marginal dollar for drugs that are not the cheapest (Huskamp, et al., 2000). Pharmaceutical manufacturers would be paid at their bid prices. In effect, the resulting insurance is a lump-sum transfer for specific drugs, with the amount of the lump sum set through a bidding process.102

But traditional Medicare has avoided the use of bidding arrangements and

differential pricing to consumers that favors low-bidding suppliers, perhaps because of the tradition of freedom of choice of provider with which it began and the resistance from the affected providers to change. Rather, it has used administered price systems, such as the PPS, with minimal or no difference among alternative suppliers in prices to the beneficiary. For example, the amount beneficiaries pay for hospital care is completely independent of the hospital they use, and for practical purposes this is true for physician services as well.103 There is no reason in principle why prescription drugs could not be an exception, but this runs counter to the political pressure to cover services from all or almost all potential suppliers.

Although price competition among pharmaceutical manufacturers can be effective

when there are competing drugs, some branded drugs have no close substitutes. It appears to me that if Medicare covers such drugs (i.e., there is no other drug in the class), there must be some kind of price control, because Medicare cannot agree to reimburse any price a manufacturer names in its bid. If a manufacturer knew Medicare would reimburse it irrespective of price, it could in principle bid an extraordinarily high price. In private insurance the PBM can negotiate with the pharmaceutical firm and can potentially simply not cover the drug if the price is too high. It is not clear that Medicare in practice could exclude the drug. Not surprisingly the pharmaceutical industry remains strongly opposed to any element of price control, arguing correctly that the monopoly rents on a few blockbuster drugs support the industry’s research and development effort (Scherer, 2000). Indeed, the industry’s fear of price controls has been an important obstacle to prior efforts at Medicare prescription drug coverage.

Medicare does now in fact spend about $2 billion per year to cover certain

outpatient drugs, and its procurement of those drugs does not inspire confidence in its

102 An important element of judgment or administered pricing comes in, however, through the determination of which drugs constitute a class. For example, all anti-depressants could be grouped as one class of drugs, or Selective Serotonin Reuptake Inhibitors (e.g., Prozac) could be one class and older trycyclic anti-depressants could be another class. Under the broader grouping, the Selective Serotonin Reuptake Inhibitors would be more expensive, since they (as of this writing) are still under patent. There could, of course, be some copayment, say $5, even for the reference drug. 103 Over 90 percent of physician bills are assigned, meaning the physician agrees to accept the Medicare fee schedule payment as payment in full. Even if the physician does not accept assignment, he or she is only permitted to charge 10 percent above the fee schedule. There is a modest experiment in bidding for durable medical equipment in Polk County, Florida, but a demonstration of competitive pricing among health plans failed to get off the ground (Dowd, et al., 2000), (Nichols and Reischauer, 2000).

47

ability to operate an administered price scheme for drugs. For those suffering from End Stage Renal Disease Medicare covers erythropoetin, a product to stimulate red blood cell production. It appears that the rate HCFA pays dialysis centers for the erythropoetin they dispense has been well above the price that the centers pay for the drug. Indeed, the margin appears large enough for the centers to offset losses they incur on other services, because the composite rate centers receive for a session of dialysis has been approximately constant in nominal terms since 1983, a fall of one-third in real terms (Medicare Payment Advisory Commission, 2001a).104

Medicare also covers certain cancer chemotherapy drugs. It has been reimbursing

those drugs at 95 percent of the average wholesale price, a price that, as in the erythropoetin case, appears to be well above the transaction prices at which oncologists actually purchase the drugs.105

No legislation resulted from the Administration’s 1999 proposal to cover drugs, in

part because of industry opposition and in part because Congressional Democrats were happy to run on this issue in 2000 if they could not obtain their preferred outcome of a universal drug benefit. Both candidates for President in 2000 made enacting a Medicare prescription drug benefit a high priority, as did many candidates for Congress. Nonetheless, the prior experiences at adding such a benefit are sobering for those like me who think such a benefit would be good policy, and much disagreement remains on the issues just discussed.

The Administration and Governance of Medicare One view of HCFA, typically found among Republicans, is a hidebound agency,

committed to the preservation of traditional Medicare. Adherents of this view frequently want to take the Medicare+Choice program out of HCFA and create a new agency to administer that part of the program (Institute of Medicine, 1996). Some adherents of this view would go even further and create an independent board to administer all of Medicare, with HCFA perhaps continuing to administer the traditional program under the supervision of the board. Many adherents of this latter view prefer to have the Congress less involved in the management of the program (they would say micro-management of the program) than it now is. A recent example of the level of Congressional management is that when HCFA proposed to use transaction prices as the basis of reimbursement for the cancer chemotherapy drugs

104 The profit margin offers an incentive to use an intravenous method of administering erythropoetin, because that method requires a higher average dose than administering the drug subcutaneously and profit rises with dose. Probably for this reason the intravenous method of administration persists despite the National Kidney Foundation’s guideline that advocates subcutaneous administration. The Veterans Administration found the necessary dose with subcutaneous administration was one-third lower. There are also markups on injectable vitamin D analogues that are covered for ESRD patients; 80 percent of hemodialysis patients (who receive dialysis at a center) prescribed this drug receive it as an injection, but nearly all peritoneal dialysis patients (who do not receive the drug at a center) receive it orally; the cost of the intravenous formulations is four to eight times greater (Medicare Payment Advisory Commission, 2001a). 105 See HCFA Program Memorandum AB-00-86.

48

described in the previous section, oncologists protested and the Congress legislated a moratorium pending a study by the General Accounting Office.106

A radically different view is that HCFA has been performing admirably under

almost impossible conditions. On this view the Congress has steadily increased the workload on HCFA, especially in what it asked in the BBA, but it has not matched the increased workload with increased administrative resources. Moreover, until the BBAs changes in HMO reimbursement, the Medicare+Choice program was expanding smartly, suggesting no bias on the part of HCFA. Adherents of this view believe the agency has performed remarkably well in this past decade given the resources at its disposal.

My own opinion is that HCFA certainly appears starved for resources (Butler, et

al., 1999), but that some organizational changes hold promise. I take these two issues in turn.

Administrative Resources. Although supporters of traditional Medicare often

point to its low administrative costs, in my view they are too low. They were probably too low even in the 1980s, but the changes that have been made in the program over the 1990s have substantially added to HCFAs administrative load.

There are many areas in which the lack of resources manifests itself. One such

area is the accuracy of the data used as the basis for the administered price systems. Those data in several instances are very old; in other instances they are unaudited. For example, the current payment system for dialysis relies on cost reports from 1977-1979, and cost reports from ESRD providers were unaudited between 1991 and 2000, when HCFA undertook an audit of 1996 cost reports.107 As another example, the hospital wage index is used to adjust payment for skilled nursing facilities and home health agencies, although they use a different mix of labor than hospitals.108 The data used to set wages across areas do not correspond to labor market areas (e.g., non-metropolitan areas of a state are considered one labor market), necessitating a cumbersome appeals process, whereby hospitals can ask to be reclassified to other geographic areas for the purposes of the wage index.109 Furthermore, wage data used to determine relative wages across areas are four years old.

Organizational Issues. A key issue with respect to an independent board is the

degree of independence that the Congress is willing to afford such a board. In a program as large and as politically sensitive as Medicare, it is not clear that Congress

106 The oncologists did not deny the markups but claimed they compensated for inadequate reimbursement for providing chemotherapy. See AMNews, October 2, 2000. 107 As one indicator of a problem, the costs reported by hospital-based dialysis centers far exceed those reported by freestanding facilities, although there is no evidence that the outpatients they treat are sicker or more costly (Medicare Payment Advisory Commission, 1998b). 108 This complaint may be a bit tendentious, since wage indices for these providers were not important before the BBA when reimbursement was cost based. 109 There are several problems with the wage index, including its not being a proper index, since it does not hold occupational mix constant. See (Medicare Payment Advisory Commission, 2001a).

49

would be willing to be substantially less involved than it is now. If not, there is probably little gain from a board.

Even short of establishing a board, the Medicare program in my view would be

well served by shifting some functions out of HCFA to other parts of the department (Etheredge, 2000), (Wilensky, 2001). These functions include Medicaid and the State Child Health Insurance Program (S-CHIP); although HCFA was originally created in the Carter Administration to bring together the administration of Medicare and Medicaid, it is not obvious that there is much synergy between them. Other functions, such as the survey and certification of nursing homes, the conditions of participation for hospitals, and the certification of clinical laboratories, could also usefully be moved out of HCFA, possibly to the CDC or the FDA. This would allow HCFA to concentrate its scarce management resources on the Medicare program.

The Congress has its own set of management issues. First, there has been a

steady increase in partisanship in the Congress, perhaps reflecting the more nearly equal division of power between the parties in the 1990s. This has made legislation more difficult. Second, the division of jurisdiction over Medicare between two committees in the House is not helpful. Third, although there are some outstanding individual staffers, in my view the expertise of Congressional staff on the key committees of jurisdiction, taken in the aggregate, has declined over time, whereas the complexity of the program has grown.110

Conclusion

I have argued that the main substantive changes in Medicare during the Clinton

years were those enacted in the Balanced Budget Act of 1997. By reducing the rates Medicare paid providers and by ending cost reimbursement, the Act generated an unprecedented slowdown in the rate of growth of Medicare spending. As a result, it importantly contributed to the Administration’s great economic achievement of the first budget surpluses in decades. It also considerably prolonged the expected life of the Part A Trust Fund, something that surely helped the Administration maintain political support among the elderly.

Much of the recent public discussion of Medicare has focused on how it should be

financed after 2010 and especially after 2020. Although this issue is certainly important, the administered price methods that Medicare is now using have serious problems and are likely to need large-scale revision over the next several years. Within traditional Medicare the separate “silo” method of paying by provider or site of service does not appear likely to work well for two reasons. First, the same service is reimbursed at different rates in different sites. Second, in the case of physician services, there is an expenditure cap that implicitly presumes the same proportion of services continue to be delivered in the office, as opposed to the outpatient department or the ambulatory surgery center. One way to reduce the silo problem,

110 Although I have not compiled any statistics, I believe the average tenure of staff on these committees is now less than in prior decades.

50

which I favor, is to bundle payment for post-acute care with the payment for inpatient hospital care, but that does not appear politically possible now. A second is to abandon the cap on spending for physician services and update fees using methods similar to those used for hospitals and other institutional providers. How Medicare should pay for new procedures and products is another of the many important pricing problems within traditional Medicare, which constitutes 85 percent of the program.

The administered price methods for Medicare+Choice also face serious issues,

especially the effort to bring more equality to rates across regions, which has unbalanced local markets between Medicare+Choice and traditional Medicare. Further, the availability of the private fee-for-service option in the floor counties potentially involves greater Medicare expenditure for very little gain to beneficiaries.

The short-run pricing issues, however, should not obscure Medicare’s long-run

financing problem, assuming historical rates of increase in health care spending resume. On the assumption that much of the historical increase in spending represents additional medical capabilities that are worth their costs, and that those capabilities will keep appearing, the issue will be how the burden of paying for these capabilities should be divided between future taxpayers and beneficiaries. That is certainly a political issue of the first rank.

By any standard Medicare a large program with politically important

constituencies; indeed, it is the largest health insurance program in the world. It now consumes more than two percent of GDP, a figure that will rise into the four to six percent range two to three decades hence. Perforce it will almost surely be on the agenda of every subsequent Congress.

51

Table 1

Supplemental Insurance Status of Medicare Beneficiaries, 1999

Supplemental Insurance Status Percent Medicare only (no supplemental coverage) 9

Medigap 27 Employer-provided supplemental 36

Medicaid 11 HMO (Medicare + Choice) 17

Source: (Rice and Bernstein, 1999). Those with employer and individually purchased coverage are classified with employer-provided.

52

Table 2 Substitution of Post Acute Care for Medicare

Inpatient Hospital Days

Yeara Inpatient Days per 1000

Beneficiaries

Skilled Nursing Facility

Days per 1000 Beneficiaries

Home Health Visits per 1000 Beneficiaries

Rehabilitation Admissions

per 1000 Beneficiaries

1981 3,827 1982 3,889 1983 3,786 1984 3,217 1985 2,823 1986 2,784 268 1,106 2.8 1987 2,815 229 1,104 3.3 1988 2,804 334 1,104 3.7 1989 2,721 889 1,350 4.0 1990 2,749 749 2,052 5.1 1991 2,728 669 2,880 6.0 1992 2,642 812 3,763 6.6 1993 2,474 948 4,661 7.2 1994 2,436 1,006 6,020 7.8 1995 2,317 1,053 7,125 8.8 1996 2,056 1,053 7,546 1997 1,979 1,519 7,519 1998 1,895 1,527 4,590

Average Annual Growth Rateb

-4.1% 15.6% 12.6%c 12.1%

Sources: Inpatient Days through 1993, Health Care Financing Review, “Statistical Supplement, 1996,” Table 23. Inpatient Days, 1994 and 1995, Statistical Abstract of the United States, 1997, pages 115-6. 1996-1998 inpatient days from http://www.hcfa.gov/stats/stats.htm. Other values calculated from Prospective Payment Assessment Commission, “Medicare and the American Health Care System,” June 1997, chapter 4. SNF values for 1997 and 1998 and home health value for 1997 are unpublished data from the Health Care Financing Administration. 1996-1998 data for rehabilitation admissions are not available. aCalendar year for hospital days through 1993; fiscal year for other values. 1994 value from Statistical Abstract because 1994 value in Statistical Supplement excludes managed care enrollees and so is biased upward. bValue is calculated from initial year shown in the Table to final year.

53

cValue through 1997 is 20.5 percent. The sharp decline in visits in 1998 reflects some undetermined mix of greater anti-fraud enforcement efforts and changes in payment that were effective in October 1997.

54

Table 3 Length of Stay, Medicare Inpatient and Total Hospital Margins

Year Length of Stay (days) Medicare

Inpatient Margin (%)a

Total Margin (%)

Update Factor (%)

All Patients Medicare Patients

1981 7.2 10.4 b not available b 1982 7.2 10.2 b not available b 1983 7.0 9.8 b not available b 1984 6.7 8.9 13.4 7.3 b 1985 6.5 8.6 13.0 6.6 not available 1986 6.6 8.7 8.7 4.3 not available 1987 6.6 8.9 5.9 3.6 not available 1988 6.6 8.9 2.7 3.5 not available 1989 6.6 8.9 0.3 3.6 3.3 1990 6.6 8.8 -1.5 3.6 4.7 1991 6.5 8.6 -2.4 4.4 3.4 1992 6.4 8.4 -0.9 4.3 3.0 1993 6.2 8.0 1.3 4.4 2.7 1994 6.0 7.5 5.6 5.0 2.0 1995 5.7 7.0 11.1 5.8 2.0 1996 5.6 6.5 15.9 6.1 1.5 1997 5.4 6.2 16.9 5.9 2.0 1998 5.3 6.1 13.7 4.3 0.0 1999 5.2 6.0 12.0 2.8 1.1 a Excludes graduate medical education payments. Including these payments would raise the margins. b Not applicable because of cost reimbursement. Sources: All patient length of stay through 1996: Prospective Payment Assessment Commission, “Medicare and the American Health Care System: Report to the Congress,” June 1997, page 89; 1997-1999 calculated from Medicare Payment Advisory Commission, “Report to the Congress,” March 2001, Table B-1. Medicare length of stay through 1996: Health Care Financing Review: Medicare and Medicaid Statistical Supplement, 1998, page 206; 1997-1999 calculated from Medicare Payment Advisory Commission, “Report to the Congress,” March 2001, Table B-1. Margins to 1993: Medicare Payment Advisory Commission, “Report to the Congress: Medicare Payment Policy,” March 1999, pages 53 and 55. 1993 and later: Medicare Payment Advisory Commission, “Report to the Congress,” March 2001, Tables B-4 and B-18. The Medicare inpatient margin is the ratio of Medicare revenue to the allocated cost of Medicare cases; the total margin is the ratio of hospital revenue from

55

all payers to total hospital cost. Operating updates from Medicare Payment Advisory Commission, “Report to the Congress,” March 2001, Table B-1.

56

Table 4 Standardized Extra Benefits

as a Function of Plan Payment, 1996

Decile Plan Payment Index Standardized Extra Benefits

US Average 1.00 $77

10 1.29 121 9 1.15 86 8 1.09 80 7 1.06 86 6 1.03 92 5 0.99 78 4 0.94 68 3 0.88 57 2 0.82 53 1 0.75 48

Source: Prospective Payment Assessment Commission, “Medicare and the

American Health Care System,” June 1997, Table 2-8. Plans are grouped in deciles of equal numbers of plans according to the level of the AAPCC. The value of extra benefits is the actuarial value of any waived premium for non-covered services and reduced cost sharing, divided by the hospital wage index for the area.

57

Table 5

The AAPCC in the Six Highest and Six Lowest Counties, 1997

The Six Highest Counties

County State Annual Rate Richmond NY $9,208 Dade FL 8,979 Bronx NY 8,739 Plaquemines LA 8,733 St. Bernard LA 8,638 New York NY 8,557

The Six Lowest Counties

County State Annual Rate

Arthur NB $2,651 Banner NB 2,656

Holmes OH 2,700 Chippewa MN 2,728 Presidio TX 2,756

Saline NB 2,773

58

Table 6 Instability in AAPCC Rates, Percent Change in 1997 Rates Relative

to 1996 Rates

Rural Counties with Large Changes

County State % Change, 1996-7 Culberson TX +37 Refugio TX +33

Logan WV +29 Gilliam OR -17 Delta CO -24

Loving TX -40

Rate Changes among Large Metropolitan Areas

County State % Change, 1996-7 Los Angeles CA 6.0

Maricopa AZ 3.8 Dade FL 8.9

Wayne MI 1.8 New York NY -0.2 Middlesex MA 7.0

___________________________________________________________________ Source: Medicare Payment Advisory Commission, unpublished materials.

59

Table 7 Payment Rates within the Washington, DC Metropolitan Area, 1997

County Annual Rate

Prince Georges, MD $7,224 Washington, DC 7,008

Montgomery, MD 5,904 Arlington, VA 5,412 Fairfax, VA 4,812

Source: Medicare Payment Advisory Commission, unpublished materials.

60

Table 8

Coinsurance for Selected Outpatient Services

APCa/Service Amount ($) Beneficiary % Share of Outpatient Coinsurance

600 Low-level Clinic Visits 10 20% 1.8% 325 Group Psychotherapy 20 26 0.8 246 Cataract Procedures 624 47 8.5 with Intraocular Lens Insert 260 Level I Plain Film 22 56 5.0 284 Magnetic Resonance 257 65 4.1 Imaging 283 Level II CAT Scan 179 74 11.5 99 Electrocardiograms 15 78 1.3 a APC is Ambulatory Patient Classification Source: (Medicare Payment Advisory Commission, 2001a), page 145.

61

Table 9

Numbers of House Staff, by year

Yeara Number of Residents Percent International Graduates

First Year Total First Year Total

1980 18,702 61,465 21 20 1985 19,168 75,518 14 17 1990 18,322 82,902 19 18 1991 19,497 86,217 24 20 1992 19,794 89,368 25 20 1993 21,849 97,370 27 23 1994 21,949 97,832 27 24 1995 21,372 98,035 26 25 1996 21,394 98,076 25 25 1997 21,808 98,043 24 26 1998 21,732 97,383 24 26 1999 22,320 97,989 26 26

aYear is the academic year beginning in year shown; e.g., 1996 is academic year 96-97.

Source: 1980-1992, Physician Payment Review Commission, “Annual Report, 1997” (page 352). 1993-1998, Rebecca S. Miller, Marvin R. Dunn, and Thomas Richter, “Graduate Medical Education, 1998-1999,” Journal of the American Medical Association, September 1, 1999, 282(9): 856. 1999, Sarah E. Brotherton, Frank A. Simon, and Sandra C. Tomany, “U.S. Graduate Medical Education, 1999-2000,” Journal of the American Medical Association, September 6, 2000, 284(9): 1121-1126, Tables 1 and 2 and Figure 1.

62

Table 10

Real Medicare Spending, CY1970-2000

Year Spending (billions of 2000 dollars)a

Percent Increase over prior year

Part A (billions of 2000 dollars)

Part B (billions of 2000 dollars)

1970 27.6 -- 19.4 8.1 1975 43.6 9.6b 30.9 12.6 1980 69.0 9.6b 47.9 21.1 1985 104.9 8.7b 70.2 34.6 1990 137.1 5.5b 82.8 54.4 1991 144.8 2.6 86.5 58.3 1992 158.1 9.2 99.0 59.2 1993 173.0 9.4 107.3 65.7 1994 183.6 6.1 116.4 67.2 1995 200.8 9.4 128.2 72.6 1996 214.2 6.7 138.9 75.3 1997 224.0 4.6 146.3 77.7 1998 221.0 -1.3 140.6 80.4 1999 217.3 -1.7 133.3 84.0 2000 221.8 2.1 131.1 90.7

a Total may not add because of rounding error. b Annualized rate over the prior five years. Sources: (Board of Trustees, 2001a), (Board of Trustees, 2001b). Deflated by the chain-weighted GDP deflator for the corresponding calendar year.

63

Table 11

Base Payment Rates for Selected High-Volume Ambulatory Services

Type of Service Description OPD Practice Expense ASC Surgical Upper GI endoscopy $347 $139 $425 Diagnostic colonoscopy 387 192 425 Colonoscopy with lesion 387 260 425 removal

Extract cataract, inset lens 1,287 --- 934 Radiology Chest X-ray, one view 38 21 ---- Mammography, both breasts 34 56 ---- Diagnostic Cardiovascular stress test 79 63 ----

Echo exam of heart 213 171 ---- Clinic Visit Office or Outpatient Visit, 48 23 ---- new patient Office or Outpatient Visit, 48 22 ---- established patient Source: (Medicare Payment Advisory Commission, 2000b), page 39. OPD is the hospital outpatient department, and ASC is ambulatory surgery center. The practice expense column shows the amount paid for services in the office that is intended to cover practice expenses, as opposed to the physician take home or work component, which is the amount that corresponds to the other two columns for office services.

64

Table 12 Medicare+Choice Contract Terminations

January 1999 January 2000 Terminations 45 41 Service Area Reductions 54 58 Enrollees Who Could Not Stay in Their Plans

407,000 327,000

Enrollees in Counties Where All Plans Withdrew

50,000 79,000

Source: (Medicare Payment Advisory Commission, 2000a), based on HCFA

announcements of December 8, 1998 and July 15, 1999.

65

Table 13 Likelihood of a Claim with the Given Diagnosis in 1995, Conditional

on a Claim for that Diagnosis in 1994

Diagnosis Likelihood (%)

Hypertension 59 COPD 62 Stroke 51

High Cost Diabetes 58 Dialysis 56

Quadriplegia/Paraplegia 52 Coronary Artery Disease 53

CHF 61 Dementia 59

Rheumatoid Arthritis 58 Source: (Medicare Payment Advisory Commission, 1998b), vol. II, ch.

2. COPD is Chronic Obstructive Pulmonary Disease; CHF is Congestive Heart Failure.

66

Table 14 Premiums for Individual Medigap Plans with and without Drug

Coverage, 1999

65 Year Old 75 Year Old Policy Ca Policy Ia Policy Ca Policy Ia Dallas, TX $1046 $2294 $1295 $2974 Denver, CO 974 2589 1199 3221 Los Angeles, CA

1502 3362 1820 4437

Miami, FL 1510 3428 1890 4158 Manchester, NH

917 1945 1247 2581

a Policy C does not cover drugs. Policy I covers 50% of drug spending above a $250 deductible to a $1,250 maximum expenditure, so the maximum value of the drug benefit is $500. Policy I also covers any physician fees in excess of Medicare’s reasonable charges, but these are limited to an additional 10% of physician fees, and few physicians charge additional fees. Policy I also covers up to 40 home health visits during recovery from an acute illness. Medicare beneficiaries who are homebound and need part-time or intermittent care already have this benefit; for others benefits are limited to $40 per visit. In addition, Policy C, but not Policy I, covers the Part B deductible of $100, which anyone using physician services is likely to satisfy. The actuarial value of the Part B deductible coverage in Plan C likely exceeds the actuarial value of the excess physician fee and home health visit features in Plan I, but in any event the premium differences of $1,000 to 2,000 would seem to vastly outweigh the additional benefits in Plan I, even for those spending the maximum amount on drugs. Source: (Gluck, 1999)

67

Table 15 Drug Coverage by Income Group Among the Elderly, 1998

Household Income

Share of Elderly Median Share of After-Tax Income Spent on Health Care

Share without Prescription Drug Coverage

All 100% 13% 34% <$10,000 18 37 35

$10,000-19,999 29 19 38 $20,000-29,999 18 15 32 $30,000-49,999 18 10 30

>$50,000 17 6 26 Source: (McClellan, et al., 2000).

68

Figure 1 Growth in Real Physician Expenditure, Calendar Years 1970-1997

0

10

20

30

40

50

1970

1975

1983

1985

1988

1991

1993

1995

1997

Spending onPhysicianServices(billions of1999 $)

Source: Health Care Financing Administration, “Medicare and Medicaid Statistical Supplement, 1999,” Table 55, adjusted by the GDP chain-type price index.

The VPS begins

69

Figure 2

Percentage of Medicare Beneficiaries Enrolled in HMOs

34 4

57

911

1415

16

02468

10121416

90 91 92 93 94 95 96 97 98 99

% of MedicareBeneficiariesEnrolled inHMOs

Source: Medicare Payment Advisory Commission, Report to the Congress,

March 1998, page 5, and unpublished data.

70

Figure 3

Selection: Mortality in 1998, HMO Enrollees as a Proportion of Traditional Medicare Enrollees, by Length of HMO Enrollment

0.85 0.79 0.83 0.85 0.87 0.90 0.89

0

0.2

0.4

0.6

0.8

1

All <1yr

1-2yrs

2-3yrs

3-4yrs

4-5yrs

>5yrs

Traditional Plan

HMO Length ofEnrollment

Source: (Medicare Payment Advisory Commission, 2000a).

71

Figure 4 Medicare Outlays Increased at a Rate Much Faster Than Federal

Tax Revenues

9.1

2.7

6.6

2.4

7.0

2.7

0

2

4

6

8

10

Real Annual % Increase

1970-80 1980-90 1990-96

Medicare

Federal TaxRevenue

Sources: Statistical Abstract, Economic Report of the President, Congressional Budget Office. Deflators: GDP deflator to 1990, CPI X1 90-96.

72

REFERENCES Aaron, Henry J., and Robert Reischauer, (1995), “The Medicare Reform Debate: What Is the Next Step?,” Health Affairs, 14(4): 8-30. Berndt, Ernst R., (2001), “The U.S. Pharmaceutical Industry: Why Major Growth in Times of Cost Containment?,” Health Affairs, 20(2): 100-114. Berndt, Ernst R., David M. Cutler, Richard G. Frank, et al., (2000), "Medical Care Prices and Output", Handbook of Health Economics, Anthony J. Culyer and Joseph P. Newhouse, ed.; Amsterdam: Elsevier, 1A: 120-180. Blendon, Robert J., Drew E. Altman, John Benson, et al., (1995), “The Public's View of the Future of Medicare,” Journal of the American Medical Association, 274(20): 1645-1648. Blendon, Robert J., John M. Benson, Mollyann Brodie, et al., (1997), “Voters and Health Care in the 1996 Election,” Journal of the American Medical Association, 277(15): 1253-1260. Board of Trustees, Federal Hospital Insurance Fund, (2001a), 2001 Annual Report of the Board of Trustees of the Federal Hospital Insurance Fund; Washington, DC: Government Printing Office. Board of Trustees, Federal Supplementary Insurance Fund, (2001b), 2001 Annual Report of the Board of Trustees of the Federal Supplemental Insurance Fund; Washington, DC: Government Printing Office. Brown, Randall S., Dolores Gurnick Clement, Jerrold W. Hill, et al., (1993), “Do Health Maintenance Organizations Work for Medicare?,” Health Care Financing Review, 15(1): 7-23. Butler, Stuart M., Patricia M. Danzon, Bill Gradison, et al., (1999), “Crisis Facing HCFA and Millions of Americans,” Health Affairs, 18(1): 8-10. Butler, Stuart M., and Robert E. Moffit, (1995), “The FEHBP as a Model for a New Medicare Program,” Health Affairs, 14(4): 47-61. Carter, Grace, Joseph P. Newhouse, and Daniel A. Relles, (1990), “How Much Change in the Case Mix Index Is DRG Creep?,” Journal of Health Economics, 9(4): 411-427.

73

Chassin, Mark, Robert H. Brook, Rolla Edward Park, et al., (1986), “Variations in the Use of Medical and Surgical Services by the Medicare Population,” New England Journal of Medicine, 314(5): 285-290. Congressional Budget Office, (1997), Predicting How Changes in Medicare Payment Rates Would Affect Risk-Sector Enrollment and Costs; Washington, DC: Congressional Budget Office. Cowan, Cathy A., Helen C. Lazenby, Anne B. Martin, et al., (1999), “National Health Expenditures, 1998,” Health Care Financing Review, 21(2): 165-210. Cutler, David M., (1995), “Cutting Costs and Improving Health: Making Reform Work,” Health Affairs, 14(1): 161-172. Cutler, David M., and Sarah J. Reber, (1998), “Paying for Health Insurance: The Tradeoff Between Competition and Adverse Selection,” Quarterly Journal of Economics, 113(2): 433-466. Davis, Margaret, John Poisal, George Chulis, et al., (1999), “Prescription Drug Coverage, Utilization, and Spending Among Medicare Beneficiaries,” Health Policy, 18(1): 231-243. Dowd, Bryan, Robert Coulam, and Roger Feldman, (2000), “A Tale of Four Cities: Medicare Reform and Competitive Pricing,” Health Affairs, 19(5): 9-29. Eggers, Paul, (1980), “Risk Differential between Medicare Beneficiaries Enrolled and Not Enrolled in an HMO,” Health Care Financing Review, 1(3): 91-99. Eggers, Paul, and Ronald Prihoda, (1982), “Pre-enrollment Reimbursement Patterns of Medicare Beneficiaries Enrolled in 'At-Risk' HMOs,” Health Care Financing Review, 4(1): 55-73. Ellis, Randall P., Gregory C. Pope, Lisa I. Iezzoni, et al., (1996), “Diagnosis-Based Risk Adjustment for Medicare Capitation Payments,” Health Care Financing Review, 17(3): 101-128. Ellis, Randall P., and Thomas G. McGuire, (1986), “Provider Behavior under Prospective Reimbursement,” Journal of Health Economics, 5(2): 129-151. Ellis, Randall P., and Thomas G. McGuire, (1993), “Supply-Side and Demand-Side Cost Sharing in Health Care,” Journal of Economic Perspectives, 7(4): 135-151. Epstein, Arnold M., Robert S. Stern, and Joel S. Weissman, (1990), “Do the Poor Cost More? A Multihospital Study of Patients' Socioeconomic Status and Use of Hospital Resources,” New England Journal of Medicine, 322(16): 1122-1128.

74

Etheredge, Lynn, (2000), “Medicare's Governance and Structure,” Health Affairs, 19(5): 60-71. Evans, Robert G., (1974), "Supplier Induced Demand: Some Empirical Evidence and Implications", The Economics of Health and Medical Care, Mark Perlman, ed.: Macmillan: 162-173. Frank, Richard G., (2001), “Prescription Drug Prices: Why Do Some Pay More Than Others Do?,” Health Affairs, 20(2): 115-128. Fuchs, Victor R., (2001), "The Financial Problems of the Elderly: A Holistic Approach"; Cambridge: National Bureau of Economic Research. Glazer, Jacob, and Thomas G. McGuire, (2000), “Optimal Risk Adjustment in Markets with Adverse Selection: An Application to Managed Care,” American Economic Review, 90(4): 1055-1071. Gluck, Michael E., (1999), "A Medicare Prescription Drug Benefit"; Washington, DC: National Academy of Social Insurance. Health Care Financing Administration, (1999), Health Care Financing Review, Statistical Supplement; Washington, DC: Government Printing Office. Holmstrom, Bengt, and Paul Milgrom, (1991), “Multitask Principal-Agent Analyses: Incentive Contracts, Asset Ownership, and Job Design,” Journal of Law, Economics, and Organization, 7((Special)): 24-52. Hsiao, William C., Peter Braun, Daniel L. Dunn, et al., (1992a), “An Overview of the Development and Refinement of the RBRVS -- The Foundation for Reform of US Physician Payment,” Medical Care, 30(11, Supplement): NS1-NS12. Hsiao, William C., Peter Braun, Daniel L. Dunn, et al., (1988), “Resource Based Relative Values: An Overview,” Journal of the American Medical Association, 260(16): 2347-2353. Hsiao, William C., Peter Braun, Edmund R. Becker, et al., (1992b), “Results and Impacts of the Resource-Based Relative Value Scale,” Medical Care, 30(11, Supplement): NS61-NS79. Huskamp, Haiden A., Meredith Rosenthal, Richard G. Frank, et al., (2000), “The Medicare Prescription Drug Benefit: How Will the Game Be Played?,” Health Affairs, 19(2): 8-23. Institute of Medicine, (1996), Improving the Medicare Market: Adding Choice and Protections; Washington, DC: National Academy Press.

75

Kane, Nancy M., and Paul D. Manoukian, (1989), “The Effect of the Medicare Prospective Payment System on the Adoption of New Technology--The Case of Cochlear Implants,” New England Journal of Medicine, 321(20): 1378-1383. Kemper, Peter, (1988), “The Evaluation of the National Long Term Care Demonstration: Overview of the Findings,” Health Services Research, 23(1): 161-174. Kessel, Reuben A., (1958), “Price Discrimination in Medicine,” Journal of Law and Economics, 1: 20-53. Kominski, Gerald F., and Stephen H. Long, (1997), “Medicare's Disproportionate Share Adjustment and the Cost of Low-Income Patients,” Journal of Health Economics, 16(2): 177-190. Laffont, Jean-Jacques, and Jean Tirole, (1993), A Theory of Incentives in Procurement and Regulation; Cambridge: MIT Press. Lubitz, James, and Gerald F. Riley, (1993), “Trends in Medicare Payments in the Last Year of Life,” New England Journal of Medicine, 328(15): 1092-1096. McClellan, Mark, (1997), “Hospital Reimbursement Incentives: An Empirical Analysis,” Journal of Economics and Management Strategy, 6(1): 91-128. McClellan, Mark, Ian D. Spatz, and Stacie Carney, (2000), “Designing a Medicare Prescription Drug Benefit: Issues, Obstacles, and Opportunities,” Health Affairs, 19(2): 26-41. McGuire, Thomas G., (2000), "Physician Agency", Handbook of Health Economics, Anthony J. Culyer and Joseph P. Newhouse, ed.; Amsterdam: Elsevier, 1A: 461-536. McGuire, Thomas G., and Mark V. Pauly, (1991), “Physician Response to Fee Changes with Multiple Payers,” Journal of Health Economics, 10(4): 385-410. Medicare Payment Advisory Commission, (1998a), Report to the Congress: Context for a Changing Medicare Program, June 1998; Washington, DC: Medicare Payment Advisory Commission. Medicare Payment Advisory Commission, (1998b), Report to the Congress: Medicare Payment Policy, March 1998; Washington, DC: Medicare Payment Advisory Commission.

76

Medicare Payment Advisory Commission, (1999a), Report to the Congress: Medicare Payment Policy, March 1999; Washington, DC: Medicare Payment Advisory Commission. Medicare Payment Advisory Commission, (1999b), Report to the Congress: Selected Medicare Issues, June 1999; Washington, DC: Medicare Payment Advisory Commission. Medicare Payment Advisory Commission, (2000a), Report to the Congress: Medicare Payment Policy, March 2000; Washington, DC: Medicare Payment Advisory Commission. Medicare Payment Advisory Commission, (2000b), Report to the Congress: Selected Medicare Issues, June 2000; Washington, DC: Medicare Payment Advisory Commission. Medicare Payment Advisory Commission, (2001a), Report to the Congress: Medicare Payment Policy, March 2001; Washington, DC: Medicare Payment Advisory Commission. Medicare Payment Advisory Commission, (2001b), Report to the Congress: Selected Medicare Issues, June 2001; Washington, DC: Medicare Payment Advisory Commission. Newhouse, Joseph P., (1986), “Rate Adjusters for Medicare Under Capitation,” Health Care Financing Review, 7(Supplement): 45-55. Newhouse, Joseph P., (1996), “Reimbursing Health Plans and Health Providers: Selection versus Efficiency in Production,” Journal of Economic Literature, 34(3): 1236-1263. Newhouse, Joseph P., (1998), “Risk Adjustment: Where Are We Now?,” Inquiry, 35(Summer): 122-131. Newhouse, Joseph P., (2001), “Medical Care Price Indices: Problems and Opportunities,” Academia Economic Papers, 29(1): 1-65. Newhouse, Joseph P., (2002), Pricing the Priceless: Medical Care, Insurance, and Capitation; Cambridge: MIT Press. Newhouse, Joseph P., Melinda Beeuwkes Buntin, and John D. Chapman, (1997), “Risk Adjustment and Medicare: Taking a Closer Look,” Health Affairs, 16(5): 26-43. Newhouse, Joseph P., and Gail R. Wilensky, (2001), “Paying for Graduate Medical Education: The Debate Goes On,” Health Affairs, 20(2): 136-147.

77

Newhouse, Joseph P., Jr. Willard G. Manning, Emmett B. Keeler, et al., (1989), “Adjusting Capitation Rates Using Objective Health Measures and Prior Utilization`,” Health Care Financing Review, 10(3): 41-54. Newhouse, Joseph P., Albert P. Williams, Bruce Bennett, et al., (1982), “Does the Geographical Distribution of Physicians Reflect Market Failure?,” Bell Journal of Economics, 13(2): 493-505. Nichols, Len M., and Robert D. Reischauer, (2000), “Who Really Wants Price Competition in Medicare Managed Care?,” Health Affairs, 19(5): 30-48. Pettengill, Julian, and James Vertrees, (1982), “Reliability and Validity in Hospital Case Mix Measurement,” Health Care Financing Review, 4(2): 101-128. Physician Payment Review Commission, (1990), Annual Report to Congress, 1990; Washington, DC: Physician Payment Review Commission. Physician Payment Review Commission, (1993), Annual Report to Congress, 1993; Washington, DC: Physician Payment Review Commission. Physician Payment Review Commission, (1996), Annual Report to Congress, 1996; Washington, DC: Physician Payment Review Commission. Prospective Payment Assessment Commission, (1996), Medicare and the American Health Care System: Report to the Congress, June 1996; Washington, DC: Prospective Payment Assessment Commission. Prospective Payment Assessment Commission, (1997), Medicare and the American Health Care System: Report to the Congress, June 1997; Washington, DC: Prospective Payment Assessment Commission. Rice, Thomas H., and Jill Bernstein, (1999), "Supplemental Health Insurance for Medicare Beneficiaries"; Washington, DC: National Academy of Social Insurance (http://www.nasi.org/Medicare/Briefs/medbr6.htm). Riley, Gerald, Cynthia Tudor, Yen-Pin Chiang, et al., (1996), “Health Status of Medicare Enrollees in HMOs and Fee-for-Service in 1994,” Health Care Financing Review, 17(4): 65-76. Scherer, F.M., (2000), "The Pharmaceutical Industry", Handbook of Health Economics, Anthony J. Culyer and Joseph P. Newhouse, eds.; Amsterdam: North-Holland, 1b: 1297-1336. Shleifer, Andrei, (1985), “Yardstick Competition,” RAND Journal of Economics, 16(3): 319-327.

78

U.S. Bureau of the Census, (2000), Statistical Abstract of the United States; Washington, DC: Government Printing Office. U.S. House of Representatives, Committee on Ways and Means, (1996), 1996 Green Book; Washington, DC: Government Printing Office. van de Ven, Wynand P.M.M., and Randall P. Ellis, (2000), "Risk Adjustment in Competitive Health Plan Markets", Handbook of Health Economics, Anthony J. Culyer and Joseph P. Newhouse, eds.; Amsterdam: Elsevier, 1A: 755-846. Wilensky, Gail R., (2001), "Regulatory Relief, HCFA Restructuring, and Medicare Reform". Subcommittee on Health of the Committee on Ways and Means, US House of Representatives,; Washington, DC. Wilensky, Gail R., and Joseph P. Newhouse, (1999), “Medicare: What's Right? What's Wrong? What's Next?,” Health Affairs, 18(1): 92-106. Wise, David A., and Richard G. Woodbury, (1994), "Policy Toward the Aged", American Economic Policy in the 1980s, Martin Feldstein, ed.; Chicago: University of Chicago Press: 741-780. Yip, Winnie, (1998), “Physician Responses to Medical Fee Reductions: Changes in the Volume and Intensity of Supply of Coronary Artery Bypass Graft (CABG) Surgeries in the Medicare and Private Sectors,” Journal of Health Economics, 17(6): 675-700. Zuckerman, Stephen, Stephen A. Norton, and Diana Verrilli, (1998), “Price Controls and Medicare spending,” Medical Care and Research Review, 55(4): 457-478.