|
TRENDSTrends In Medical Spending By Age, 1963–2000
We combine household surveys and total spending data to analyze trends in medical spending from 1963 to 2000. During this nearly forty-year period, spending grew fastest among the elderly. Per person spending among the elderly grew rapidly from 1963 to 1987, but this trend then reversed during the next decade, reflecting reforms to Medicare’s physician and hospital payment systems. Prescription drug spending growth accelerated throughout the time period; it grew fastest among the nonelderly, for whom prescription drug insurance coverage is most extensive.
Understanding the growth of medical spending is essential in an aging society. Spending has increased for all age groups in the past half-century, but the differential across age groups is important. A 1 percent faster growth in medical costs for the aged, for example, would raise Medicare spending by 64 percent in fifty years. Similarly, a prescription drug benefit for Medicare that misestimated growth rates by 1 percent per year would cost 10.5 percent more one decade out than currently envisioned. There are many reasons to expect growth in medical costs to differ between the elderly and nonelderly. On one hand, the elderly are more likely to be insured and to use more services. On the other hand, evidence suggests that disability is falling among the elderly but declining slowly or even increasing among the non-elderly. This trend would reduce relative spending on the elderly over time.1 To date, there has been relatively little examination of the differential growth of spending by age.2 Medicare’s actuaries, for example, assume that costs for the elderly will rise in the future at the same rate as those of the non-elderly.3 Consistent with this, researchers have shown that Medicare spending rose at about the same rate as private insurance spending during the past thirty years.4 But Medicare accounts for only about half of medical care spending on behalf of the elderly. Thus, this comparison is not wholly definitive about the likely spending burden of an expanded Medicare program that includes prescription drugs, a Medicaid program that covers nursing home care, and other possible changes to social insurance programs. Comparisons across different time periods are also of great interest. Until the 1980s no insurance policies had serious cost containment mechanisms in place. The growth of spending in recent periods may therefore differ from the historical experience. If spending growth among the elderly slowed after the early 1980s, for example, that would support the superiority of the federal government’s cost containment efforts over efforts in the private sector. The reverse would indicate the advantage of increased managed care enrollment. In this paper we examine the growth of total medical spending by age and time, and we decompose that spending growth into types of services and sources of increase.5
To estimate per person medical spending by age group and service category, we used national household surveys combined with published data on total spending by service category and payer. We briefly outline our methods below and provide a more detailed description in an online technical appendix.6 Our measure of medical spending includes hospital care, physician and clinical services, prescription drugs, home health care (after 1977), nursing home care, and other services (dental care, durable medical equipment, and other professional services). Although other types of spending, such as spending on health clubs, might contribute to health, we exclude such measures from our estimates. An ideal measure of health spending might remove all spending that has little direct effect on health. Our measures exclude administrative costs of insurance, measuring instead payments to providers of medical goods and services. However, our measures of spending on nursing home care implicitly include spending on housing. Medical spending surveys. We used data from five national household surveys: the 1963 and 1970 Surveys of Health Services Utilization and Expenditures from 1963 and 1970 (N = 7,759 and 11,619, respectively); the 1977 National Medical Care Utilization and Expenditure Survey (N = 38,815); the 1987 National Medical Expenditure Survey (NMES) (N = 34,459); and the 2000 Medical Expenditure Panel Survey (MEPS) (N = 23,652). We used the "household component" of each survey, which contains spending and utilization data for a nationally representative sample of the civilian, noninstitutionalized population.7 Using the expenditure surveys, for each service category we estimated the fraction of spending accounted for by different age groups. Details of our adjustment to the expenditure surveys are provided in the technical appendix.8 Allocation of spending totals. The Centers for Medicare and Medicaid Services (CMS) National Health Accounts (NHA) provide data on total spending by service category and payer for every year since 1960.9 The spending totals in these accounts do not exactly match the spending totals calculated from the national surveys, for several reasons. First, the surveys do not include nursing home spending or spending on institutionalized people. Second, the national surveys tend to underreport spending on services such as lab tests and physician services, because of difficulties obtaining complete payment information.10 Third, the national surveys in earlier years record charges ("list price") and in later years record payments made for services provided; the NHA records total payments received by providers. Fourth, the national surveys exclude payments to providers that are not linked to the provision of specific services, such as Medicare direct graduate medical education payments. Because the NHA totals provide a consistent measure of total payments received by providers, we used them to calculate total spending by service category and payer. Within each category, we calculated spending totals separately for Medicare and all other payers. We allocated spending across age groups within each service category using our estimates of the age distribution of spending from the national surveys; we then allocated spending across age groups within each service category. These methods generate estimates of spending by age group while addressing the potential downward bias in the surveys discussed above. These methods also adjust for differences that arise from reporting charges versus costs. A team of researchers recently attempted to reconcile the total spending amounts for 1996 obtained from the NHA and MEPS.11 To make the NHA totals more consistent with the national surveys, we made several adjustments to NHA spending using methods that were similar in spirit to those outlined by Thomas Selden and colleagues: that is, we excluded philanthropy, and we shifted spending on hospital-based nursing homes and home health agencies from the hospital care category to the nursing home category and home health category, respectively.12 Because the national surveys exclude nursing home residents, we allocated total nursing home spending based on National Center for Health Statistics (NCHS) data on the nursing home population. Given the absence of detailed data on nursing home expenditures by age over time, we assumed that nursing home spending per resident was the same for all age groups. To estimate spending by nursing home residents on service categories other than nursing home care, we assumed that within ten-year age groups, spending per person among residents was twice that of nonresidents.13 All survey-based estimates used sample weights corrected so that the sum of the sample weights within each age group equaled the U.S. population within that age group based on data from the U.S. Bureau of the Census.14 Computing per person spending and growth rates. Using the spending allocation method described above and population data from the Census Bureau, we estimated per person spending for each year, age group, and service category and annual growth rates in spending. We also estimated the share of spending accounted for by each age group and per person spending by age group relative to people ages 35–44. We used the gross domestic product (GDP) deflator to inflate spending to 2002 dollars, although changes in relative spending by age are similar regardless of the deflator used.15
Growth in spending by age. Between 1963 and 2000, per person spending grew within all age groups, but growth was slower after 1970 (Exhibit 1  ). Patterns of rapid relative spending growth among the elderly were not constant throughout the study period, however. Between 1987 and 1996, relative spending growth slowed greatly among those age sixty-five and older. After 1996, annual spending growth accelerated to 5.8 percent for the 65–74 age group, while spending actually decreased 1.2 percent per year among those age seventy-five and older.
Relative spending by age group. Exhibit 2 documents the trend in "relative spending," defined as the ratio of per person spending within an age group to spending within the 35–44 age group. Two important trends appear. First, in the long run, relative spending became much more skewed toward the elderly; second, the trend of faster spending growth among the elderly reversed after 1987.
The elderly’s share of spending. From 1963 to 2000, the elderly accounted for a slightly larger share of the population, but their share of spending nearly doubled (Exhibit 3 ). Since 1987, their share of spending has dropped slightly.
Spending growth by age and service category. Exhibit 4 illustrates several interesting differences in spending trends by age group and service category. First, consider the trend in spending growth on the two most costly service categories, hospital care and physician and clinical services. Spending on these services grew more rapidly among the elderly than among the nonelderly during 1963–1987 but then slowed to 0.8 percent growth per year for the elderly’s hospital care and 1.9 percent per year for their physician and clinical services during 1987–1996. While growth in hospital spending was slow for both age groups during 1996–2000, spending for the elderly’s physician and clinical services accelerated to 5.1 percent per year between 1996 and 2000, in contrast to the more modest rate of 2.6 percent among the nonelderly.
Second, home health care spending trends differ dramatically by age and time period. Following the Balanced Budget Act (BBA) of 1997, Medicare payments to home health agencies fell dramatically, reflected in the 12.5 percent annual decline in elderly spending on home health during 1996–2000.
Third, unlike all other service categories, growth in per person spending for prescription drugs accelerated throughout the period (Exhibit 4
Over the period 1963–2000, per person spending grew most rapidly among those age sixty-five and older. This resulted in a rising share of spending attributable to the elderly, beyond what we would expect from the aging of the population. The long view masks two distinct trends. From 1963 to 1987, spending growth among the elderly was high relative to that of the nonelderly; since then, spending per person has risen more rapidly among the nonelderly. This reversal resulted from several factors, including changes in the elderly growth rate for major service categories and the more rapid growth of prescription drug spending since 1987. Our results closely mirror those found in earlier estimates of spending by age using less complete measures of medical spending.16 As did earlier work, we find that spending grew faster among the elderly compared with other groups through 1987, but by the mid-1990s this trend had abated.17 We confirm that population aging accounts for only a small part of medical spending growth since 1970: only 0.2 percentage points of the annual rate of 4.3 percent.18 Patterns of spending among the elderly compared with the nonelderly correspond with major changes in Medicare policy during the past three decades. Consistent with other recent research that omitted the non-Medicare portion of spending among the elderly, we find that from 1987 to 1996, hospital spending growth for the elderly was below that for the nonelderly, as was that of spending for physician services.19 This likely reflects the implementation of Medicare’s prospective payment system (PPS) for inpatient short-stay hospital care in 1983 and physician fee schedule in 1992.20 The presence of managed care in Medicare, which peaked in the late 1990s, also brought estimated cost savings for both managed care and traditional Medicare beneficiaries.21 While these forces temporarily slowed spending on acute care services among the elderly, the major driver of spending growth, technological innovation, likely will continue to exert upward pressure on spending trends.22 As further evidence of Medicare’s influence, home health care spending for the elderly rose sharply in the late 1980s and early 1990s and then declined just as abruptly following passage of the BBA, which overhauled Medicare payments for home health care. The BBA also changed payment policy for skilled nursing home care, altering provider incentives toward less costly care. After a period of double-digit growth rates between 1987 and 1996, average spending on skilled nursing care grew only 1 percent per year between 1996 and 2000. The BBA clearly contributed to the reversal in spending trends among the elderly.23 Our findings offer insight on spending implications of the recent legislation adding a Medicare prescription dr ug benefit. Unlike other service categories, prescription drug spending growth has accelerated in every period and among all age groups since 1963. Despite the current lack of universal insurance coverage for prescription drugs, spending among the elderly on prescription drugs grew much faster than other medical spending. The $395 billion Congressional Budget Office (CBO) cost estimates of the Medicare Prescription Drug, Improvement, and Modernization Act (MMA) of 2003 assumed that drug spending would grow at the same rate as other Medicare-covered services. To understand the importance of this assumption, consider what would have happened between 1987 and 2000 if drug spending for the elderly had risen at the same rate as other Medicare-covered services. Prescription drug spending would have grown from $537 per person in 1987 to $628 in 2000; instead, it grew to $1,254 per person in 2000, nearly double the rate of the hypothetical scenario. In January 2004 the White House estimated that the new Medicare legislation would cost $534 billion, much more than the CBO’s estimate, and the discrepancy does not stem from differences in drug spending growth.24 This example illustrates the pitfalls that arise when trends in medical spending by age and service category are ignored. Despite our gloomy conclusions regarding the cost of Medicare drug coverage, our findings suggest that when cost containment efforts are in place, Medicare can influence spending among the elderly. Spending patterns among the elderly mirror Medicare’s cost containment efforts, with dramatic declines in spending growth for targeted services. Overall, Medicare policymakers have demonstrated an ability to contain costs without widespread evidence of harmful effects on health.
Ellen Meara (meara{at}hcp.med.harvard.edu) is an assistant professor of health economics in the Department of Health Care Policy, Harvard Medical School, in Boston, Massachusetts, and a faculty research fellow at the National Bureau of Economic Research (NBER), in Cambridge. Chapin White is a postdoctoral fellow at the NBER. David Cutler is a professor of economics in the Harvard University Department of Economics and a research associate at the NBER. Chapin White acknowledges the National Institute on Aging (NIA) (Grant no. T32 AG00186) and the Agency for Healthcare Research and Quality (AHRQ) for providing pre- and postdoctoral training grants during the preparation of this paper. David Cutler acknowledges research support from the NIA (Grant no. R01 AG15911). The authors are grateful to Joel Cohen, Alan Monheit, and Sam Zuvekas of AHRQ for clarifying technical details of the Medical Expenditure Panel Survey and its predecessors during the early stages of the study. They also received helpful input from Sean Keehan and Aaron Catlin of the Centers for Medicare and Medicaid Services, Office of the Actuary.
This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Top
Methods
Study Results
