Medicare Beneficiaries’ Costs Of Care In The Last Year Of Life
Abstract
This paper profiles Medicare beneficiaries’ costs for care in the last year of life. About one-quarter of Medicare outlays are for the last year of life, unchanged from twenty years ago. Costs reflect care for multiple severe illnesses typically present near death. Thirty-eight percent of beneficiaries have some nursing home stay in the year of their death; hospice is now used by half of Medicare cancer decedents and 19 percent of Medicare decedents overall. African Americans have much higher end-of-life costs than others have, an unexpected finding in light of their generally lower health care spending.
End-of-life costs are only slightly higher for persons who died than for survivors with similar characteristics.
As baby boomers age and face the deaths of parents, relatives, and friends, care of the dying is becoming a topic of widespread personal importance. There is a growing interest in examining the current state of care and identifying opportunities for improving care at the end of life. U.S. policy on the financing of acute care at the end of life is, almost by default, Medicare policy. In 1997 roughly 2.3 million American adults died.1 More than 80 percent of these were covered by Medicare, with about 1.75million over age sixty-five and another 0.15 million entitled to Medicare on the basis of disability or end-stage renal disease (ESRD).2
This Data Watch summarizes and extends findings from a recent study of Medicare beneficiaries’ costs and use of care in the last year of life.3 It quantifies the costs, disease burden, cognitive impairment, nursing home use, hospice use, and site of death of Medicare decedents in the mid-1990s.
Data And Methods
We used three main sources of patient-level data to characterize Medicare decedents. First, Medicare claims and eligibility data for a 0.1 percent random sample of beneficiaries for 1993 through 1998 provide detailed data on spending in the last twelve months of life (for decedents) or calendar year (for survivors, those not dying in a given calendar year). This sample captures about 8,000 deaths. Spending data were adjusted for geographic differences in Medicare price levels and for inflation. Second, Medicare Current Beneficiary Survey (MCBS) Cost and Use files for 1992 through 1996 provide data on spending by all payers in the calendar year of death, as well as information on residence, disability, and sociodemographics. This captures about 3,500 deaths. Third, the 1993 National Mortality Followback Survey (NMFS), preliminary version, provides information from a sample of death certificates including about 8,000 persons age sixty-five and older matched to survey responses from next of kin. Except where noted, information is based on all aged (sixty-five and older) and disabled (under age sixty-five) fee-for-service Medicare beneficiaries from the Medicare data sources.
We found substantial agreement across sources on aggregate estimates of key factors such as site of death and nursing home use. Cost differences by race were similar whether race was taken from administrative data (claims) or self-reported survey data (MCBS). The major disagreement across sources was lower use of hospice reported in the NMFS, attributable to the age of the NMFS data (1993) and to rapid growth of hospice over the study period. Differences discussed in the text passed standard t-tests of statistical significance (p < .05), accounting for design effects of the clustered or stratified surveys (MCBS, NMFS). Cost data were estimated in dollars without log or other transformation.
Estimated prevalence of disease was based on diagnoses from all Medicare claims during the given year. Compared with more restrictive approaches, this gives a more plausible estimate of prevalence of key diseases, including dementia.4 Dementia, which is important in characterizing the health condition of the elderly, historically has been underreported in administrative data.5 Any claims-based estimates of disease prevalence may somewhat under- or overstate true prevalence because of incomplete diagnosis coding or erroneous or “rule-out” diagnoses on some claims.
We constructed an estimated “cause of death” variable based on physician spending data for Medicare fee-for-service beneficiaries. (Cause of death from death certificates is not routinely matched to Medicare administrative data.) This variable identifies the main reason for health care use, defined as the disease accounting for the plurality of the physician care dollars in the last year of life. Using this approach, cancer decedents appear to be identified well, but other diseases may not be. Only the cancer/noncancer contrast is used here.6
This is a retrospective analysis of persons who died. Other than for cancer decedents, the timing of death is typically highly unpredictable. Except for those in hospice (where certification of short expected lifespan is required), there should be no inference that care was delivered in anticipation of death.
Study Findings
Demographics.
About 5 percent of Medicare beneficiaries die each year. This decedent population is much older than are those who did not die in a given calendar year and is composed of a somewhat lower proportion of women (Exhibit 1⇓).
Demographics, Medicare Spending, And Prevalence Of Disease For Medicare Decedents And Survivors, 1993–1998
Medicare costs.
The share of Medicare spending for persons in the last year of life has been stable for two decades. For the mid-1990s decedents’ per capita Medicare program outlays were about six times higher than survivors’ (Exhibit 1⇑). This ratio is slightly lower than a similar estimate for 1979.7 After minor adjustments for comparability with earlier estimates, spending in the last year of life accounted for 27.4 percent of all Medicare outlays for the elderly, similar to the 26.9–30.6 percent range in earlier decades.8
Medicare is not the sole source of funds for this population. The MCBS captures payments from all sources (including out-of-pocket costs) and so provides an estimate of Medicare’s share of all health care spending. Based on MCBS data for the calendar year of death (on average, the last six months of life, accounting for about 70 percent of cost for the last twelvemonths), Medicare paid 61 percent of decedents’ costs, Medicaid paid 10 percent, and other payers paid 12 percent. Out-of-pocket costs accounted for 18 percent.9 Because nursing home spending is a major component of out-of-pocket costs, the oldest old had the highest out-of-pocket spending.
Disease prevalence.
Decedents typically have multiple significant medical problems in the year of death. Diagnoses from Medicare claims data were used to identify any mention of conditions among the leading causes of death in the elderly. Using this measure of illness, the typical Medicare decedent had almost four significant diseases in the last year of life, while the average for survivors was slightly more than one in the typical calendar year. About three-quarters of decedents had some mention of heart disease in the claims data (with about half having mention of heart failure); roughly one-third had cancer, stroke, chronic obstructive pulmonary disease, or pneumonia/influenza. More than one-fourth had some form of dementia.
Decedents’ high end-of-life costs are largely a consequence of this substantial disease burden and are only modestly higher than costs of survivors with similar characteristics. Decedents’ costs were less than 50 percent higher than those of a matched cohort of survivors, when the match was based on age and diagnoses, and less than 30 percent higher when the match was based on age, diagnoses, and some hospitalization during the year.
Decedents were far more likely to have had some episode of nursing home care (as evidenced by some claim for a physician visit in a nursing facility). Coverage by Medicaid was much higher for decedents than for survivors (Exhibit 1⇑).
Elderly decedents differ greatly by age (Exhibit 2⇓). Those ages sixty-five to seventy-four were more likely to be men, to die of cancer, and to have high costs. As age at death increased, so did prevalence of dementia, nursing home use, and Medicaid coverage, while cancer deaths and Medicare per capita outlays fell. For decedents age eighty-five or older, most were women, most had some nursing home episode, and more than 40 percent had a diagnosis of dementia recorded in claims in the last year of life.
Characteristics Of Elderly Medicare Decedents, By Age, 1993–1998
Hospice use.
The greatest change over the study period was the increased use of hospice. The proportion of decedents using hospice rose from 11 percent in 1994 to 19 percent in 1998 (Exhibit 3⇓). By the end of the study period (1998) more than half of Medicare cancer decedents had used hospice, but only 10 percent of others had done so. Medicare+Choice (M+C) enrollees who died were much more likely than others were to have used hospice. On average, over the 1993–1998 period, 25 percent of M+C decedents had used hospice, versus 15 percent of FFS enrollees who died.
Percentage Of Medicare Decedents With Some Use Of Hospice In Last Year Of Life, Selected Years 1994–1998
Race and poverty.
One of the most surprising findings was that end-of-life costs were higher for minorities and for those living in ZIP code areas with high poverty rates. Medicare per capita spending for minority decedents was about 28 percent higher than for others, while spending for beneficiaries in high-poverty ZIP codes was 43 percent higher than in ZIP codes with the lowest poverty rates (Exhibit 4⇓).
Last-Year-Of-Life Costs, By Race And Area Income, 1993–1998
In part, the higher costs were attributable to case-mix—about 7 percent of minority Medicare decedents had had very costly deaths from ESRD, versus less than 2 percent of the remainder of the Medicare population. Yet after ESRD deaths were excluded, costs for minority decedents remained more than 20 percent above costs for others. Analysis of self-reported race data from the MCBS show that costs were elevated only for African Americans, not others.
Facility residence and site of death.
Both administrative and survey data show that nearly 40 percent of beneficiaries had some nursing home stay in their last year of life. This includes a mix of long and short stays. Interviews with next of kin in the NMFS show that 22 percent of elderly decedents were full-time nursing home residents, while another 16 percent were part-year residents in the year of death (Exhibit 5⇓).10
Residence And Site Of Death For The Elderly, 1993
Beneficiaries’ likely place of death was strongly associated with place of residence and use of hospice care. About two-thirds of full-year facility residents and half of part-year residents died in a nursing home. Death at home was typical only for hospice enrollees.
Interpretation And Suggestions For Policy
In the late 1980s and early 1990s analysts questioned whether “the high cost of dying” was a major driver of the growth of health care spending.11 The “no” answer reached then still appears to be true. Medicare outlays for the last year of life continued to be a stable fraction of total Medicare spending through 1998.
Perhaps we should instead ask why decedents’ costs are such a stable fraction of Medicare outlays. We suggest a simple answer. Much of what has been labeled the “high cost of dying” is just the cost of caring for severe illness and functional impairment. Decedents’ costs are, roughly speaking, not much different from those of others with similarly complex medical needs.
In recent years much has been done to consider quality issues specific to end-of-life care. The results of this study, by contrast, emphasize the need to include end-of-life issues in discussions of other aspects of Medicare policy. First, efforts to introduce greater coordination of care for the frail and chronically ill in traditional Medicare will necessarily affect end-of-life care. Decedents are drawn mainly from the frailest strata of the Medicare population. New initiatives for the frail elderly could acknowledge high expected mortality rates by including end-of-life care in program benefit design.
Second, nursing homes play a key role. We found that 38 percent of Medicare decedents (and the majority of decedents over age eighty-five) spent some time in a nursing home in the last year of life. Primarily, this means that discussion of coherent Medicare end-of-life policy should include coordination of acute and long-term care benefits and the interface between Medicare and Medicaid. At the same time, a more widespread popular understanding of how many Americans will likely end their days in a nursing home might increase interest in improving options and funding in this area.
Third, for beneficiaries not in a nursing home, not in a managed care plan, and not dying of cancer, the health care system offers little support for end-of-life care other than through traditional acute care providers. Surveys suggest that about half of persons with a serious chronic illness would prefer to die at home.12 Hospice provides organized support for in-home death, but hospice remains typical only for those dying of cancer. Recently adopted guidelines clarify hospice admission criteria for noncancer patients, but the lack of accurately predictable life span means that hospice (as now implemented in Medicare) is unlikely to be an option formost.13 This seems an important and potentially costly gap if it encourages inpatient deaths. It is not clear whether future solutions, if any, would focus on modification of the existing hospice benefit or on expansion of new options centered on care of the chronically ill.
Looking at the broader picture, discussion of Medicare restructuring should logically include end-of-life care. Almost all Americans enroll in Medicare for the rest of their lives. The data suggest that most of us will pass through a period of substantial illness burden, functional impairment, or cognitive impairment prior to death. In our own self-interest, we should judge a proposed future Medicare system, at least in part, on the likelihood that it will provide good care to persons at the end of life.
Footnotes
-
Christopher Hogan is president of Direct Research LLC in Vienna, Virginia. June Lunney is a research consultant at RAND, in Arlington, Virginia. Jon Gabel is vice-president, Health System Studies, at Health Research and Educational Trust in Washington, D.C. Joanne Lynn is director of RAND’s Center to Improve Care of the Dying, also in Arlington.
-
This work was funded by the Agency for Healthcare Research and Quality, Grant no. R01-HC10561-01, with support from the Medicare Payment Advisory Commission. The authors thank Jeff McCartney and Tom Bell of Social and Scientific Systems Inc. for help in preparing the data. Helpful comments from two anonymous referees are gratefully acknowledged.
