four years (through 2001), all three groups spent nearly twice as much on Medicare-covered services as beneficiaries in the reference group.

Those methods are highly stylized and conceptual illustrations. In practice, which groups to identify and which methods to use would depend on the actual intervention strategies that might be implemented. Moreover, the extent to which those targeted beneficiaries reduced their spending would ultimately rest on the ability to devise and implement effective intervention strategies, clinical or otherwise, to change beneficiaries' use of medical services.

The Concentration of Medicare
Expenditures

Medicare spending is highly concentrated, with a small
number of beneficiaries accounting for a large proportion
of the program's annual expenditures. In 2001, the costli-
est 5
percent of beneficiaries enrolled in Medicare's fee-

for-service (FFS) sector accounted for 43 percent of total spending, while the costliest 25 percent (defined as the high-cost group in this paper) accounted for fully 85 percent of spending (see Figure 1). In this context, spending includes expenditures paid for by all parties—including the Medicare program itself, beneficiaries, and thirdparty payers such as medigap insurers-for all services covered by the Medicare program.2 (For a description and discussion of the Medicare data used in this analysis, see Box 1.) Real (inflation-adjusted) spending among the most expensive 5 percent of beneficiaries averaged about $63,000 per person in 2001, with the least expensive person in that group spending more than $35,400 (see Table 1). Among the most expensive 25 percent of beneficiaries, spending averaged about $24,800, with the least expensive beneficiary in that group spending over $6,200 in 2001. By contrast, the least expensive 50 percent of Medicare beneficiaries accounted for only 4 percent of total spending, with costs in 2001 averaging about $550 per person.

The concentration of Medicare spending has lessened slightly since the early 1990s. From 1991 to 2001, there was a large increase in the level of Medicare spending, with total annual Medicare expenditures per FFS beneficiary growing by more than 40 percent in inflationadjusted terms, from $5,080 to $7,310.3 However, the rate of increase in spending was larger among low-cost beneficiaries than among high-cost ones. On average, real per capita spending among the bottom 75 percent of beneficiaries grew at 6.8 percent per year over that 10year period, whereas spending among the top 25 percent

2. Spending by the Medicare program itself is more concentrated. For example, the most expensive 5 percent of Medicare FFS beneficiaries accounted for 48 percent of annual Medicare FFS spending in 2002, and the most expensive 25 percent accounted for 88 percent. See Medicare Payment Advisory Commission, Report to the Congress: New Approaches in Medicare (June 2004), Figure 2-1. 3. The increase in national health expenditures during that period was even larger, with an average annual growth rate of inflationadjusted spending per person of over 5 percent between 1990 and 2001. See www.cms.hhs.gov/statistics/nhe/historical/tl.asp. Part of the difference in the rates of spending growth is attributable to the fact that Medicare did not cover outpatient prescription drugs-a fast-growing component of national health expendirures during that period. In addition, the Balanced Budget Act of 1997 reduced payments for various Medicare-covered services.

HIGH-COST MEDICARE BENEFICIARIES

Box 1.

Methodology of This Analysis

The Congressional Budget Office's (CBO's) analysis of Medicare expenditures presented in this paper is based on longitudinal data of Medicare claims from 1989 through 2001 for a 5 percent sample of Medicare beneficiaries enrolled in the fee-for-service sector of the Medicare program. The sample was derived from claims records maintained by the Centers for Medicare and Medicaid Services.

The data contain information on the enrollment and entitlement status of each beneficiary, his or her demographic characteristics, and monthly expenditures for all Medicare-covered services (short-term hospital, other hospital, skilled nursing facility, outpatient, physician, home health, hospice, and durable medical equipment). In this paper, total expenditures for Medicare-covered services include third-party payments and beneficiaries' share of payments through deductibles and copayments, as well as the amounts paid by the Medicare program. CBO converted all expenditures into 2005 dollars using the GDP deflator. (Those constant-dollar expenditures can be interpreted as the opportunity cost of health care.) Although the data contain some diagnostic information reported in claims files (such as beneficiaries' diagnosis-related group, or DRG), they generally lack detailed clinical information. Moreover, because outpatient prescription drugs were not covered by Medicare during the years of the sample, the database also does not include spending for prescription drugs.

To be included in a given year of data, a beneficiary had to have at least one month of enrollment in both Part A (Hospital Insurance) and Part B (Supplementary Medical Insurance) of the Medicare program. Beneficiaries enrolled in managed care were excluded from the analysis because their expenditure information was not available. The resulting sample contains approximately 1.6 million beneficiaries per year.

The number of admissions to hospitals and skilled nursing facilities was constructed from inpatient and skilled nursing claims. The number of visits to physicians' offices and emergency departments was constructed from codes (according to the Healthcare Common Procedure Coding System, or HCPCS) reported in physicians' claims.

Seven chronic conditions were considered in the analysis: asthma, chronic obstructive pulmonary disease, chronic renal failure, congestive heart failure, coronary artery disease, diabetes, and senility. A beneficiary was defined as having a chronic condition if he or she had the diagnosis reported in physicians' claims data (as the primary or secondary diagnosis) for at least one month in a given year. Detailed diagnosis codes were grouped into general categories using the Clinical Classification Software developed by the Agency for Healthcare Research and Quality.

4

of beneficiaries grew at 3.3 percent per year. As a result, the share of Medicare spending by the top one-quarter of beneficiaries decreased from 88.8 percent in 1991 to 85.0 percent in 2001 (see Table 1).

4. Those trends are consistent with the fact that medical spending for physician-provided care (used by both low-cost and high-cost Medicare beneficiaries) grew faster during that period than did spending for hospital-provided care (used largely by high-cost beneficiaries).

Factors Affecting the Degree of Concentration A high degree of concentration of expenditures is not unique to the Medicare population. Health care expenditures in the general population show similar patterns." In fact, they are even more concentrated: in 1996, for example, the costliest 5 percent of the U.S. population accounted for 55 percent of total health care spending.

5. See Marc L. Berk and Alan C. Monheit, "The Concentration of Health Care Expenditures, Revisited," Health Affairs, vol. 20, no. 2 (March/April 2001), pp. 9-18.

HIGH-COST MEDICARE BENEFICIARIES

Table 1.

n.a. not applicable.

That skewed distribution of medical spending is rooted in the fundamental reason that people value insurance. Events that people typically insure against, like flood or fire, involve a small probability of a very expensive outcome. Similarly, in health care, although individuals may know about their need for medical services to some degree, their exact amount of spending on medical care is variable and unpredictable. For instance, most people do not know whether they will have a heart attack, even if they are fully aware of their relative risk factors. Health insurance spreads the financial risks of adverse health outcomes across the insured population, so that the small fraction of people who incur very high expenses of severe illness are financially protected. Both the probability of adverse health outcomes and the expense of medical care to treat them affect the degree of concentration of spending.

To the extent that the probability and the nature of ill health vary across subgroups of the Medicare population, one would expect to see varying degrees of concentration in spending across those groups. A striking example is the very small group of beneficiaries with end-stage renal disease (ESRD), who have chronic kidney failure and require dialysis or kidney transplantation. Most people with that condition have very high medical spending. As

a result, that group has a much more even distribution of expenditures across its members than does the larger Medicare population. For example, the most expensive 5 percent of ESRD patients accounted for only 17.4 percent of spending by all ESRD patients in 2001; in comparison, the most expensive 5 percent of all Medicare FFS beneficiaries accounted for 43 percent of spending by all Medicare beneficiaries (see Table 2). Similarly, beneficiaries with chronic medical conditions have high average annual medical spending, but it is also more evenly spread across that group than is spending for the overall Medicare population.7

Table 2.

HIGH-COST MEDICARE BENEFICIARIES

The prevalence of chronic conditions, which typically require ongoing care and treatment to maintain health and functional status and to slow the progression of the disease, was also strongly linked to high expenditures and the use of medical resources. More than 75 percent of high-cost beneficiaries were diagnosed with one or more of seven major chronic conditions in 2001. More than 40 percent of high-cost beneficiaries had coronary artery disease, and about 30 percent had each of three other conditions-diabetes, congestive heart failure, and chronic obstructive pulmonary disease. All of those conditions were much less prevalent among low-cost beneficiaries.

In terms of the medical services they received, the highand low-cost groups were similar in that they both visited physicians regularly (see Table 4). The vast majority of

Medicare beneficiaries in both groups saw a physician in 2001; however, among high-cost beneficiaries who visited a physician, the average number of visits during the year was 11, compared with six visits among low-cost beneficiaries who visited a physician at least once. High-cost beneficiaries were also much more likely to have been admitted to a hospital or a skilled nursing facility than were members of the low-cost group and to have been treated in a hospital emergency room during the year.

The Persistence of Medicare
Expenditures

If the goal of policymakers is to ultimately direct intervention strategies toward high-cost beneficiaries and change their use of Medicare services, it is important to consider patterns in Medicare spending over relatively long periods of time, not just over one year. Do individuals who make heavy demands on the Medicare program one year continue to do so in subsequent years? Or are

HIGH-COST MEDICARE BENEFICIARIES

Table 3.

the high-cost beneficiaries changing each year? If there is high turnover among high-cost beneficiaries, intervention strategies designed to change their use of Medicare services could be difficult to implement successfully because the time available to affect their spending may be limited.

Expenditure Patterns Over Time

The transition of Medicare beneficiaries between highand low-cost status in two successive years is illustrated in Table 5. For Medicare beneficiaries who were high cost in 1997, nearly half (44 percent) were also in the high-cost category the next year, compared with one in six (17 percent) of low-cost beneficiaries. If the transition between cost categories was purely random, 25 percent of the survivors in each group would have been expected to be high cost in the second year.

sistence of their high-cost status (see Figure 2). As discussed above (and indicated by the darkest bars in Figure 2), 44 percent of high-cost beneficiaries in 1997 had large Medicare spending again in 1998. That fraction dropped off in subsequent years, nearly reaching 25 percent four years later, in 2001. A similar spending pattern preceded high-cost beneficiaries' 1997 experience: nearly half of those who would be high cost in 1997 were high cost in 1996, and about one-quarter were high cost four years prior to 1997.

That pattern of spending makes intuitive sense. Although the presence of serious chronic illness is common among high spenders, many types of adverse health shocks that result in very high spending (such as a heart attack and the subsequent bypass operation) are episodic and largely random. It is less likely that a person would have a series