|
|
| |||||||||||||||||||||||||||||||
Background Increased attention is being paid to the amount and types of medical services rendered in the period before death. There is a popular impression that a greater share of resources is being devoted to dying patients than in the past. We examined trends in the proportion of Medicare expenditures for persons 65 years old or older in their last year of life to determine whether there were any changes from 1976 to 1988.
Methods Using Medicare program data for 1976, 1980, 1985, and 1988, we classified Medicare payments according to whether they were made for people in their last year of life (decedents) or for survivors. We also assigned expenses for care in the last year of life according to intervals of 30 days before the person's death and examined trends according to age.
Results Reflecting the large overall increase in Medicare spending, Medicare costs for decedents rose from $3,488 per person-year in 1976 to $13,316 in 1988. However, Medicare payments for decedents as a percentage of the total Medicare budget changed little, fluctuating between 27.2 and 30.6 percent during the study period. Payments for care during the last 60 days of life expressed as a percentage of payments for the last year also held steady at about 52 percent. Furthermore, the pattern of lower payments for older as compared with younger decedents also prevailed throughout the study period.
Conclusions The same forces that have acted to increase overall Medicare expenditures have affected care for both decedents and survivors. There is no evidence that persons in the last year of life account for a larger share of Medicare expenditures than in earlier years.
The purpose of this study was to examine the trend in the proportion of all Medicare payments that is accounted for by people in the last year of life, to determine whether the proportion has changed. Since the publication of the 1984 study10 there have been two major changes in the Medicare program that may affect the amount of Medicare expenditures for those in the last year of life -- the prospective payment system for hospitals and the hospice benefit. In addition, there have been changes in the attitudes of society toward care for the dying, such as increased attention to following the wishes of the terminally ill as expressed in advance directives16.
Although data on Medicare payments for people in the last year of life cannot be used to identify which expenses were for patients who were clearly terminally ill, monitoring these payments can help identify trends in the care of beneficiaries before death. If payments for people in their last year increased more rapidly than overall Medicare payments, more attention to services received in the last months of life might be called for.
Methods
The data used in these analyses came from the Continuous Medicare History Sample, a longitudinal file on a 5 percent random sample of Medicare beneficiaries. The data on numbers of beneficiaries and dollars were multiplied by 20 to estimate totals. The file is part of the Medicare Statistical System, which compiles information from Medicare claims submitted by physicians, hospitals, and other providers. New beneficiaries are added to the Continuous Medicare History Sample, and the records of beneficiaries who die are retained in the file. Information on the use of Medicare services and dates of death is periodically added to the records of the beneficiaries.
We obtained comparable data for 1976, 1980, 1985, and 1988 on Medicare payments made on behalf of people in the last year of life. These years were chosen because 1976 was the earliest year for which data were available, and 1988 the latest, at the time the study began. Not all the tables show data for 1980 and 1985, but in no case would the data for those years have altered the conclusions.
Our sample included only Medicare beneficiaries 65 years of age or older; enrollees under the age of 65 who were entitled to Medicare because of disability were not included. The sample was restricted to Medicare beneficiaries not enrolled in health maintenance organizations (HMOs), because the records of use of Medicare services are incomplete for HMO members. The number of persons in our sample was 1,167,966 for 1976, 1,258,702 for 1980, 1,362,099 for 1985, and 1,455,424 for 1988.
The data are limited to payments for services covered by Medicare. Medicare covers hospital inpatient services, skilled-nursing services provided in a qualified skilled-nursing facility after hospitalization, home health services, services by physicians and other medical providers and suppliers, outpatient services (including those provided by hospitals, ambulatory surgical centers, and rural clinics), and hospice care (since October 1983). Important services not covered by Medicare are nursing home care not qualifying for Medicare payment and outpatient drugs. Medicare paid for only 2 percent of nursing home care in 198817.
Assignment of Costs to Decedents and Survivors
For each study year, we assigned Medicare payments either to decedents (persons in their last year of life) or to survivors (all others). In the case of the calendar year 1988, for example, beneficiaries who survived through December 31, 1989, were identified as survivors for 1988. All their person-years of enrollment in 1988 and their Medicare payments for services provided in 1988 were assigned to survivors. Person-years of enrollment and payments in 1988 for persons who died in 1988 were assigned to decedents. In addition, a portion of the payments for calendar year 1988 for persons who died in 1989 was assigned to survivors, and a portion to decedents. Payments for services during 1988 that were received within 365 days of the day of death were assigned to decedents. Any other payments for services in 1988 were assigned to survivors.
For example, if a person died on July 30, 1989, the 211th day of the year, that person would contribute the last 154 days of 1988 (i.e., 365 minus 211) to the person-year count for 1988 decedents. All Medicare payments for services provided to that person during the last 154 days of 1988 would be assigned to decedents. The first 211 days of 1988 for the same person would be assigned to the person-year count for survivors, and all payments for services provided during the first 211 days of 1988 would be assigned to survivors.
Data for hospitals and skilled-nursing facilities were assigned to survivors or decedents on the basis of the date of discharge. For example, for a 1988 decedent, the payments for a hospital stay that began in December 1987 and ended in January 1988 would be counted entirely with 1988 decedents. All hospice payments were assigned to decedents. For other services of physicians and other medical care givers and for home health care and outpatient care, the Continuous Medicare History Sample does not record the date of service but maintains only a total of payments for the calendar year. For these services, a simple algorithm was used to apportion the 1988 payments between decedents and survivors for persons dying in 1989. In the case of people dying on July 30, 1989, a total of 154/365 of the 1988 payments for these services would be assigned to decedents, and the remaining payments to survivors.
The same methods were used for the 1976, 1980, and 1985 data. To control for changes in the composition of the study population according to age and sex and in the death rates of Medicare beneficiaries from 1976 through 1988, we adjusted the payment data for 1980, 1985, and 1988 to the 1976 Medicare population.
Estimating Monthly Costs in the Last Year of Life
To estimate payments in the last year of life according to the number of months before death, we needed to overcome the limitations of the information on dates of service in the Continuous Medicare History Sample file noted above. We first categorized decedents according to the month of death and then computed annual per capita expenses for them according to the month of death. If one assumes that deaths in January occur evenly throughout the month, decedents in January have an average of 15.5 days of enrollment before death. Thus, their calendar-year expenses represent an estimate of payments made in the last 15.5 days before their deaths. Similarly, decedents in February have an average of 45 days of enrollment before death, and therefore the calendar-year expenses on their behalf approximate the expenses in the last 45 days before death. Thus, for each month we derived a value for the average number of days of enrollment for persons dying in that month, and a value for the average expenses associated with the average number of days of enrollment. Using linearizing methods described by Tukey,18 we fitted a curve to the 12 data points to provide an estimate of expenses for any particular number of days before death. In this analysis, data are presented for months defined as periods of 30 days.
Trend in the Percentage of Decedents among High-Cost Beneficiaries
We calculated the percentage of high-cost beneficiaries (defined as those for whom payments exceeded the 95th and 99th percentiles in a 12-month period) who were decedents in each of the four study years. Because this analysis required data on individual beneficiaries for a 12-month period, we chose subsamples of the decedents who died in December of each study year, because nearly 12 months of Medicare payments were reflected in their calendar-year totals. We also randomly selected 1/12 of the beneficiaries who were alive on December 31 of each study year as a sample of survivors.
These subsamples of decedents and survivors were pooled, and the 95th and 99th percentiles of Medicare payments were computed for the pooled sample for each study year. We calculated the percentage of beneficiaries in the sample for whom the payments exceeded these percentiles and who were decedents. Some of those alive on December 31 died the next year; consequently, some of the Medicare payments on their behalf during a study year were incurred during the last year of life. We treated these persons as survivors, because our purpose was to examine the trend in the relative number of high-cost decedents rather than to derive exact estimates of costs for decedents.
Standard Errors
Approximate relative standard errors were estimated on the basis of two subsamples of persons. (The relative standard error is the standard error of the estimate divided by the estimate.) The first subsample included all persons who died in December of each study year. The second included 1/12 of those alive on December 31 of each study year. We estimated the variances of the mean expenditures for decedents and survivors directly from records for the subsample. Relative standard errors for the total expenditures were estimated on the basis of formulas developed by Dr. Thomas Jabine, formerly chief mathematical statistician of the Social Security Administration. Because of large samples and consequently small relative standard errors, confidence intervals for estimates are not shown, but in each table the largest relative standard error of any estimate is noted.
Results
Payments for Decedents and Survivors
Between 1976 and 1988, the annual number of deaths among Medicare enrollees 65 years of age or older increased from 1.22 million to 1.49 million (Table 1). This increase was commensurate with an increase in Medicare beneficiaries from 23.4 million to 29.1 million. The percentages of elderly Medicare beneficiaries who died in the four study years were very similar, ranging from 5.1 to 5.4 percent. The 1.49 million deaths of elderly beneficiaries in 1988 represented 69 percent of all deaths in the United States19.
|
Medicare payments grew considerably during the study period, from $3,488 per person-year in 1976 to $13,316 per person-year in 1988 for decedents and from $492 to $1,924 for survivors (Table 2). For decedents, inpatient hospital care accounted for over 70 percent of Medicare payments in each of the study years. Hospital care accounted for a smaller percentage of the total for survivors. For example, in 1988, inpatient hospital services accounted for 70.3 percent of all Medicare payments for decedents, as compared with 53.3 percent of payments for survivors. Payments for physicians and other medical services accounted for about one fifth of payments for decedents and about one third of payments for survivors.
|
Decedents with High Costs
Decedents made up 35 to 39 percent of the 5 percent of beneficiaries with the highest costs and 43 to 50 percent of the 1 percent of beneficiaries with the very highest costs during the study years. There was no trend over time in the percentage of high-cost beneficiaries who were decedents.
Medicare Payments According to Age
In both 1976 and 1988, Medicare payments in the last year of life generally decreased as age at death increased (Table 3). In 1988, Medicare payments for decedents 65 to 69 years of age averaged $15,436, whereas those for decedents 90 years of age or older averaged $8,888. For survivors in 1988, Medicare payments increased with age through the group that was 85 to 89 years old, after which they decreased.
|
In both 1976 and 1988, about half of all Medicare costs in the last year of life were incurred in the last 60 days of life, and about 40 percent were incurred in the last 30 days (Figure 1). The high proportion of expenses incurred in the last 30 days reflects the fact that over half the deaths among elderly persons occurred in the hospital20.
|
The principal finding of this study is that the share of Medicare expenditures accounted for by persons in their last year of life remained virtually the same from 1976 through 1988. During this period, there was considerable growth in Medicare expenditures, with payments per person-year increasing by nearly four times both among all elderly persons and among decedents. However, the proportion of total dollars spent for decedents changed little. Apparently, the same forces that have acted to increase overall Medicare expenditures -- inflation, new techniques, and greater intensity of care -- have affected care both for decedents and for survivors.
It is perhaps not surprising that the share of the Medicare budget accounted for by beneficiaries in the last year of life did not change. First, the uncertain prognoses of many severely ill patients may limit the scope of possible changes in the care of dying patients. Because our study was retrospective, we knew which beneficiaries died. In actual practice, physicians are often unsure of a patient's prognosis21. They are faced with choices among treatments for patients who are very sick, but not necessarily dying. This point is illustrated by the finding that there were similar numbers of decedents and survivors among the 1 percent of beneficiaries for whom costs were highest.
Second, we do not know how medical costs may be affected by changes in attitudes about the appropriate amount of care for the dying. For example, an evaluation of the Medicare hospice benefit concluded that the benefit has limited potential for reducing costs for the terminally ill22. In addition, only a small proportion of Medicare decedents (3 percent in 1988) received the hospice benefit.
We found that for survivors there was a substantial shift in the location of care from the inpatient to the outpatient setting. For decedents the shift was relatively slight. The shift away from the inpatient hospital setting is thought to be due to incentives created by Medicare's prospective payment system, as well as to advances in medical techniques that permit more care to be provided on an outpatient basis23. Care for less severely ill patients, such as patients undergoing cataract surgery or hernia repair, has been shifted to outpatient settings, thereby increasing the average severity of the cases treated in the hospital24. The decreasing use of inpatient care to treat survivors probably reflects this trend. It is undoubtedly less feasible to shift much of the care for decedents to outpatient settings.
The relation between age, survival status, and the use of Medicare services also changed little between 1976 and 1988. The use of acute care services declined with age for decedents; among survivors it increased until the age of 90, after which it declined. Scitovsky also found that the use of hospitals by decedents decreased with age7. This suggests that providers may be less inclined to order aggressive interventions for the very old.
The percentage of Medicare payments made during the last year of life that were incurred during the final 60 days was virtually identical in 1976 and 1988. This suggests that "heroic" efforts to preserve life in the last few months, to whatever extent they occur, have not had a disproportionate effect on the increase in Medicare costs. If such efforts had become more frequent, one might have expected an increase in health care expenses during the final month or two as compared with the rest of the last year. This was not the case. We also found no increase in the percentage of high-cost enrollees who were decedents.
Two limitations of these results should be kept in mind. First, they pertain only to Medicare-covered services. Only a small part of all nursing home care, an important service for the elderly, is covered by Medicare. Although the use of acute care services generally declines with age among decedents, the use of nursing home care increases,5,7 and overall health expenses (including nursing home costs) do not decline substantially with age at death5,7. The second limitation is that our results reflect costs in the last year of life, not the costs of terminal illness. Our data do not indicate when the decedent's terminal illness began, which in many cases would be difficult to determine anyway.
We are left with a picture of a health care system in which spending has increased both for patients who died and for patients who survived. Our data cannot address the issue of the appropriateness of care for the terminally ill. There is no evidence, however, that increased costs for persons in the final year of life are a special problem, different in magnitude from the overall growth of Medicare expenditures.
We are indebted to James Beebe for statistical advice, to James Greer for computer programming, and to Florence Beckman for assistance in the preparation of the manuscript.
Source Information
The statements contained in this report are solely those of the authors and do not necessarily reflect the views or policies of the Health Care Financing Administration.
From the Health Care Financing Administration, 6325 Security Blvd., Rm. 2504, Oak Meadows Bldg., Baltimore, MD 21207, where reprint requests should be addressed to Mr. Lubitz.
References
| |||||||||||||||||||||||||||||||
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2008 Massachusetts Medical Society. All rights reserved. |
Previous
