|
F U T U R E E L D E R L Y
C H R O N I C D I S E A S E
26 September 2005
The Lifetime Burden Of
Chronic Disease Among The Elderly
Reducing chronic illness in future
elderly cohorts will have
only modest effects on Medicare’s financial stability.
By Geoffrey F. Joyce, Emmett
B. Keeler, Baoping Shang, and
Dana P. Goldman
ABSTRACT:
The high costs of treating chronic diseases suggest that reducing their
prevalence would improve Medicare’s financial stability. In this paper
we examine the impact of selected chronic diseases on the distribution of health
spending and its variation over the course of disease. We also use a microsimulation
model to estimate these conditions’ impact on life expectancy and health
spending from age sixty-five to death. A sixty-five-year-old with a serious
chronic illness spends $1,000–$2,000 more per year on health care than
a similar adult without the condition. However, cumulative Medicare payments
are only modestly higher for the chronically ill because of their shorter life
expectancy.
Concerns over the solvency of Medicare have focused attention on the medical
expenses of elderly Americans with serious chronic conditions. Medicare spending
is highly skewed: About 10 percent of Medicare beneficiaries, almost all of
them chronically ill, account for three-fourths of program outlays each year.1
Per capita medical spending is three to ten times higher for older adults reporting
one or more chronic conditions than for those reporting none.2
Beneficiaries who have been treated for conditions such as heart attacks, cancer,
or diabetes in one year have above-average expenditures in future years as well.3
Thus, from many perspectives, any attempt to control Medicare costs must address
the demand for medical care among this subset of beneficiaries.
The high annual costs associated with many chronic illnesses suggest that reducing
their prevalence might lower lifetime health care spending. Yet because healthier
people live longer, cumulative Medicare spending might be largely invariant
to beneficiaries’ general health status at age sixty-five or seventy.4
Initial assessments of disease management programs tell a similar story. Better
coordination and management of chronically ill patients improve quality of care
but do not reduce overall treatment costs.5
In this paper we use a microsimulation model to examine how seven of the most
common chronic illnesses among the elderly affect average life expectancy and
health care spending from age sixty-five to death. For each condition we examine
health care spending by stage of disease, the distribution of medical spending,
and the impact on Medicare and Medicare beneficiaries. For example, what are
the lifetime costs of diabetes or cancer at age sixty-five to Medicare and to
Medicare beneficiaries? How do costs vary by stage of disease and by years after
diagnosis? Understanding the burden of specific chronic diseases and the persistence
of high spending in the elderly can have broad implications in designing appropriate
insurance coverage, risk-adjusting payments to health plans, and identifying
the economic consequences of screening and prevention.
Study Data And Methods
Data.
We used data, compiled for the Future Elderly Project (FEP), that link extended
Medicare claims (1991–1999) to each respondent in the Medicare Current
Beneficiary Survey (MCBS) from 1992 to 1999. The MCBS is a nationally representative
sample of the Medicare population. It was initially designed as a fixed panel
with no predetermined limit on how long sample members would be followed. However,
because of attrition, the MCBS moved to a rotating panel beginning in 1994.
Each fall approximately one-third of the sample is retired, and a new cohort
is included to meet a target sample size of approximately 12,000. The MCBS asks
about chronic conditions, use of health care services, medical care spending,
and health insurance coverage. The survey also asks respondents about limitations
in their activities of daily living (ADLs). In the MCBS, reimbursements are
categorized into nine different service groups, such as inpatient care, ambulatory
services, outpatient prescription drugs, home health, and institutional care,
and they include services not typically covered by Medicare. The MCBS includes
sample weights that reflect the probabilities of selection, oversampling, and
nonresponse.
MCBS data were linked to historical Medicare claims to track medical care use
and costs over an extended period of time. The data included up to nine years
of claims data for people age sixty-five or older in 1991, where the actual
length of the panel depends on survival. Medicare claims data included use and
cost information on inpatient hospitalizations, outpatient hospital care, physician
services, home health care, durable medical equipment, skilled nursing home
services, hospice care, and other medical services.
Information about chronic conditions was obtained in two ways. MCBS participants
were asked at each interview if they had ever been diagnosed with specific conditions
(lifetime prevalence). We also used International Classification of Diseases,
Ninth Revision (ICD-9) diagnostic codes from the Medicare claims to identify
beneficiaries who received treatment for each condition during the year (treated
prevalence). For most conditions, we used a single diagnosis to classify a patient.
However, we required two codes for several conditions in case “rule-out”
diagnoses were recorded in the claims data.6
Study sample.
The study sample includes 28,482 Medicare beneficiaries and 73,260 person-years
of data. We excluded years in which beneficiaries were under age sixty-five
or enrolled in Medicare health maintenance organizations (HMOs). We also excluded
beneficiaries residing in a facility at age sixty-five during any point in the
year.7 We chose seven high-prevalence or high-cost
conditions (or both) that were continuously asked about in the MCBS: hypertension,
diabetes, cancer (excluding skin), chronic obstructive pulmonary disease (COPD),
acute myocardial infarction (AMI), coronary heart disease (CHD), and stroke.
Our definition of cancer includes lung, breast, prostate, colon, uterine, throat,
bladder, kidney, and brain cancer. COPD is a category of respiratory illness
that includes chronic bronchitis, emphysema, and some forms of asthma.
We examined resource use by three mutually exclusive disease stages: incident,
maintenance, and terminal. We identified a terminal year based on the date of
death recorded on the claims. For nonterminal years, we defined an incident
year of treatment based on a primary diagnosis for the condition in year t
and the absence of a related diagnosis in prior years. All other years were
classified as maintenance years.
Future Elderly Model.
To simulate the consequences of having each chronic disease at age sixty-five,
we used the Future Elderly Model (FEM), a microsimulation that tracks elderly
Medicare beneficiaries over time to project their health, functional status,
and health care expenses over their remaining lifetime. We focus on the elderly
because they incur a large fraction of total health care costs and because their
expenses are borne primarily by Medicare and Medicaid.
We simulated health transitions by first predicting each person’s probability
of dying, getting a new health condition, or achieving a new functional status.
We then drew a random number to determine the simulated outcome. For each person
in the age cohort, our simulation yielded a complete lifetime profile of medical
costs, disease history, and age at death. More specifically, we predicted annual
life expectancy by estimating a set of proportional hazard models. The hazard
of getting a disease and dying depends on patient demographics (sex, education,
race, ethnicity), risk factors (obesity, smoking history), and other health
conditions. We treated all health states as absorbing—that is, once a
person got an illness, it was permanent—and we modeled transitions into
each condition. For instance, for a person age sixty-five in 1992, we predicted
his or her probability of death and probability of developing selected disease
conditions at age sixty-six. We then used a random draw to determine whether
or not a person had any of these conditions in 1993, controlling for demographic
characteristics and risk factors. The process was repeated to obtain a complete
lifetime profile of medical costs, disease history, and age at death for each
person in the sample. We then simulated disease prevalence and costs for a representative
cohort of sixty-five-year-olds.
We estimated average annual Medicare spending and total health care spending
paid for by the beneficiary and all third-party payers. Lifetime spending is
the sum of expenditures over each person’s simulated lifetime. This follows
the approach of James Lubitz and colleagues.8 Spending
was based on pooled weighted least squares regressions, with total Medicare
reimbursement and total health care reimbursement as the dependent variables
and patient demographics, risk factors, disease conditions, and ADL limitations
as the independent variables. All costs are in 1999 dollars, and future costs
are discounted at a 3 percent rate.9
We then used the simulated cohort of sixty-five-year-olds and their subsequent
spending and life expectancy to estimate the independent effects of each chronic
disease from age sixty-five to death. We regressed the simulated outcomes (life
years, total costs, Medicare costs) on patients’ characteristics and their
disease conditions at age sixty-five. We used a generalized linear model with
a logarithmic link function rather than ordinary least squares (OLS), but our
results are largely insensitive to this choice.
Study Results
Lifetime prevalence rates at age sixty-five are shown in Exhibit
1, based on participants’ responses to ever being told by a physician
that they had a specific condition. Exhibit
1 also shows the fraction of beneficiaries treated for each condition at
age sixty-five based on ICD-9 diagnostic codes (that is, annual treated prevalence
derived from medical claims). Lifetime prevalence rates are much higher than
annual treated prevalence for AMI and some cancers because of their acute nature
and the associated mortality. Many people who survive a heart attack or prostate
cancer before age sixty-five do not require ongoing medical treatment for the
condition in subsequent years. In contrast, lifetime and treated prevalence
rates are less divergent for conditions that require ongoing treatment such
as diabetes and hypertension.
A large fraction of older adults with chronic diseases have multiple chronic
conditions, as well as some type of activity limitation. Our data indicate that
only a minority of sixty-five-year-olds are free of comorbidity. Only 20 percent
of sixty-five-year-olds with hypertension or cancer are free of comorbidities
and physical limitations, and the fraction falls below 5 percent for patients
with CHD and stroke. The most common comorbid conditions among the elderly are
hypertension, osteoarthritis, and heart disease. About two-thirds of sixty-five-year-olds
with a history of AMI report having hypertension, 53 percent have osteoarthritis,
and 52 percent have heart disease. Among those with hypertension at age sixty-five,
more than 20 percent report having diabetes (not shown).
The impact of a chronic illness on the demand for health care varies considerably
by condition as well as across patients with the same condition (Exhibit
2). A typical Medicare beneficiary age sixty-five or older with hypertension
spends about $9,500 on total medical care in the incident year, and expenses
fall about 10 percent in subsequent, nonterminal years. The pattern is quite
different for cancer, stroke, and CHD. An average beneficiary with one of these
conditions spends about twice as much as a hypertensive in the incident year
($17,000–$19,000), but their costs decline 40–50 percent in subsequent
years (among survivors). The reduction in health care spending following an
incident of AMI is even more dramatic. Average spending in the incident year
exceeds $25,000 per elderly survivor but declines 60 percent in subsequent years.
Considerable attention has been devoted recently to the health care costs of
decedents as well as to the fraction of lifetime costs that occur in the final
year of life.10 We found that average health care
spending in the terminal (calendar) year is relatively insensitive to disease
condition. Average health care expenses in the last year of life average $25,000–$30,000
for all of our conditions, although we lack information on the cause of death.
Because medical spending in a given year is highly skewed, average spending
can mask the burden of disease for some patients. The median level of spending
is modest for most of the conditions, ranging from $3,000 to $10,000 annually.
However, the top 10 percent of spenders incur annual expenses that are three
to six times higher.
Prior research suggests that elderly beneficiaries with high medical spending
in one year are likely to have higher-than-average spending in other years.
Yet even among survivors, it is unusual to remain in the highest deciles of
spending for more than a few years.11 Exhibit
3 presents average medical spending in the years before and after diagnosis,
where the year of diagnosis is defined as having one or more medical claims
for a condition in calendar year t and the absence of a claim in prior
years. Health costs in the years before diagnosis are modest and fairly similar
across conditions. However, the spike in health spending in the incident year
(t) varies considerably. Incident costs are much higher for acute,
life-threatening conditions such as AMI and stroke than for more progressive
illnesses such as hypertension and diabetes. We also found that health spending
declines in the years after diagnosis but remains above the prediagnosis level
for all seven conditions.
Despite high medical costs associated with serious chronic illness, out-of-pocket
expenses are moderate for the average person. Average annual out-of-pocket costs
in an incident year range from about $1,100 to $1,875 and are fairly stable
over time. Although average costs are modest, some beneficiaries do face high
medical expenses. Exhibit
4 presents out-of-pocket spending for the ninetieth percentile of beneficiaries.
Expenses are fairly constant across disease conditions, particularly in the
incident year. Ten percent of sixty-five-year-olds spend $2,200 or more out
of pocket on health care in an incident year, with modestly higher costs for
AMI patients. Moreover, for six of seven conditions, out-of-pocket expenses
three years later are largely unchanged. The lone exception is AMI, for which
out-of-pocket expenses fall greatly three years later.
Exhibit 5 presents estimates of average life expectancy and health spending
from age sixty-five to death, by health status and for all beneficiaries. A
healthy beneficiary, defined as having no comorbidities or ADLs at age sixty-five,
lives an additional 21.7 years on average and incurs (discounted) health care
expenses of about $87,000 over his or her remaining lifetime. In comparison,
a beneficiary with three or more comorbidities at age sixty-five will live 5.3
fewer years on average and spend an additional $35,000 on medical care over
his or her remaining lifetime.
Exhibit 6 shows the impact of specific chronic diseases on life expectancy
and total medical costs. For each condition we present the average reduction
in life expectancy and change in health care spending relative to those of a
similar beneficiary without the condition. For example, having diabetes at age
sixty-five lowers average life expectancy by 3.1 years (compared with a similar
sixty-five-year-old without the condition). Diabetes also increases health care
costs by $1,930 per year and $13,032 cumulatively. Similarly, having a stroke
by age sixty-five reduces average life expectancy by three years and increases
annual treatment costs by $1,389. However, cumulative costs differ by less than
$4,000 for those with and without a history of stroke at age sixty-five, because
the reduction in life expectancy almost completely offsets the increase in annual
health care spending.
Chronic illness has a similar effect on Medicare payments, although the level
of spending is slightly lower. Annual Medicare expenses increase by about $500–$1,500
for people with a serious chronic illness at age sixty-five, while cumulative
expenses increase by $2,000–$12,000 across the seven chronic conditions.
One possible explanation of why we observe only modest differences in cumulative
spending by health status is that the incidence of many diseases increases with
age. Exhibit
7 shows average incidence rates for each condition between ages 65–69
and ages 65–74. For example, the annual incidence rate of hypertension
is 6.7 percent at ages 65–69 and rises to 6.9 percent per year if the
age range is extended to 74. Based on these rates, nearly 30 percent of those
without hypertension at age sixty-five will develop it by age seventy, and about
half will develop the condition by age seventy-five.
Discussion
Many chronic diseases affecting the elderly have considerable impact on life
expectancy and health care costs. The expected reduction in average life expectancy
ranges from 0.3 years for a sixty-five-year-old with hypertension to more than
three years for an average beneficiary with stroke or diabetes. Although all
of the conditions we studied increased annual health care costs, cumulative
spending from age sixty-five to death was only modestly higher for the chronically
ill. A person with CHD at age sixty-five will incur about $12,000 more in total
medical expenses over his or her remaining life years than in the absence of
the condition. Our findings are generally consistent with recent work by Lubitz
and colleagues, who found that a seventy-year-old in good (self-reported) health
lives longer but incurs similar medical costs over his or her remaining lifetime
to those of a seventy-year-old in poor health.12
Why does the prevalence of a chronic condition at age sixty-five have only modest
effects on cumulative spending? First, as shown in Exhibit
7, many beneficiaries without a condition at age sixty-five will develop
it in subsequent years. Thus, the cost savings from better health at age sixty-five
do not accrue indefinitely. Second, the costs incurred in the final year of
life are substantial and largely invariant to disease condition or age. Extending
life by several years reduces the high (discounted) costs incurred prior to
death, but they cannot be avoided altogether under the current system of care.
Glass half full or
half empty?
Whether one views these findings as a half-empty or half-full glass depends
on one’s perspective. Exhibit
5 indicates that being free of comorbidity and disability at age sixty-five
lowers cumulative spending in old age by more than 16 percent relative to an
“average” beneficiary. On the other hand, more policy-relevant comparisons
presented in Exhibit
6 show how life expectancy and medical spending would change if we reduced
the prevalence of each chronic condition in the entering cohort of sixty-five-year-olds.
In all cases, average life expectancy increases, but the cost savings are modest.
Our study also illustrates how cost trajectories vary across conditions. Acute,
life-threatening illnesses such as heart attack, stroke, or cancer cause a large
spike in medical costs in the incident year, largely because of the use of inpatient
care. Although health care spending declines considerably in subsequent years
(among survivors), it remains higher than before diagnosis. In contrast, more
progressive illnesses such as hypertension, diabetes, or indolent cancers lead
to more-modest cost increases in the incident year and remain near those levels
in subsequent, nonterminal years.
Our estimates of annual health care spending are not directly comparable to
prior cost-of-illness studies that rely on cross-sectional data. Those studies
typically estimate the costs of a specific condition over a year, holding constant
other comorbidities that are likely to be affected by the underlying condition
of interest. In contrast, the use of longitudinal data on a simulated cohort
of sixty-five-year-olds allows us to capture the stream of costs, both monetary
and in life years, associated with a condition, allowing each condition to influence
the likelihood of developing other illnesses.
Our findings indicate that reducing the prevalence of chronic disease at age
sixty-five will lead to only modest reductions in Medicare outlays. All seven
chronic conditions we studied increased total Medicare reimbursements. Yet the
cost savings from averting these conditions at age sixty-five would yield cumulative
savings of $500–$12,000 per case—a small percentage of the total
average spending from age sixty-five to death.13
Reducing chronic illness in future elderly cohorts might not yield large savings
to the Medicare program. Yet the combination of modest cost savings and improvements
in longevity and quality of life suggest that early prevention and screening
for some conditions could be effective public health measures. Reducing the
prevalence of chronic disease in younger populations not only might lower prevalence
rates at age sixty-five, but could also reduce the likelihood of ever developing
these conditions at older ages. These implications are particularly germane
given the rising prevalence of disability and chronic disease in pre-Medicare
populations. Between 1984 and 1996 disability rates of people ages 40–49
rose one percentage point, or almost 40 percent, and this increase coincided
with substantial growth in rates of asthma and diabetes among the young.14
Study limitations.
Our study has several limitations. First, we excluded Medicare beneficiaries
residing in nursing homes or other institutions at age sixty-five for any part
of the year. Thus, our estimates of life expectancy and costs per disease apply
only to community-dwelling Medicare beneficiaries. However, our estimates capture
all health care costs, including long-term care and Medicaid expenses, after
age sixty-five for our simulated cohort. Second, we identified the chronically
ill and defined disease stage from survey data and medical claims. The main
concern with the use of claims data is false positives if “rule-out”
diagnoses are recorded on the claims. We tried to minimize this error by requiring
two or more medical claims for diabetes and hypertension. Further, drawing our
sample of chronically ill from claims data may overstate the average severity
of illness, since patients with mild cases are less likely than those with severe
sickness to receive treatment. Finally, our analyses focus exclusively on the
elderly and ignore changes in quality of life. Reducing the prevalence of chronic
disease in younger populations can lead to much larger improvements in health-adjusted
life expectancy and reductions in health care spending.
Chronic diseases are responsible
for a majority of U.S. health care spending, and just five chronic diseases—heart
disease, cancer, stroke, COPD, and diabetes—account for more than two-thirds
of all deaths.15 Many of these conditions are often
preventable, and effective measures exist to prevent or delay much of the burden
associated with them. Although this is a goal worth pursuing, it will have only
modest effects on the financial stability of the Medicare program.
Principal funding for this study came from the Centers for Medicare and
Medicaid Services (CMS Contract no. 500-95-0056), with additional funding from
the National Institute on Aging through its support of the RAND Roybal Center
for Health Policy Simulation (P30AG024968) and the UCLA Claude D. Pepper Older
Americans Independence Center (AG16677). The authors are solely responsible
for the paper’s contents. No statement in this paper should be construed
as being an official position of the CMS.
NOTES
1. M.L. Berk and A.C. Monheit, “The Concentration of Health
Expenditures: An Update,” Health Affairs 11, no. 4 (1992): 145–149.
2. P. Fishman et al., “Chronic Care Costs in Managed Care,”
Health Affairs 16, no. 3 (1997): 239–247.
3. A.M. Garber, T.E. MaCurdy, and M.B. McClellan, “Persistence
of Medicare Expenditures among Elderly Beneficiaries,” NBER Working Paper
no. 6249 (Cambridge, Mass.: National Bureau of Economic Research, October 1997).
4. J. Lubitz et al., “Health, Life Expectancy, and Health
Care Spending among the Elderly,” New England Journal of Medicine
349, no. 11 (2003): 1048–1055.
5. B. Fireman, J. Bartlett, and J. Selby, “Can Disease
Management Reduce Health Care Costs By Improving Quality?” Health
Affairs 23, no. 6 (2004): 63–75.
6. Physicians may code for the associated signs or symptoms
of a condition rather than the condition itself. To avoid misclassifying patients
as having a condition based on these “rule-out” diagnoses, we required
beneficiaries to have two or more codes for diabetes and hypertension.
7. Starting in 1997, the MCBS changed the wording of some survey
questions regarding disease prevalence and disability for beneficiaries residing
in a facility. As result, self-reported prevalence and disability rates fell
dramatically relative to 1992–1996. Because of this inconsistency, we
excluded beneficiaries residing in a facility at age sixty-five. However, for
those living in the community at age sixty-five (the study sample), our estimates
of medical spending capture all types of service use, including skilled nursing
facilities.
8. Lubitz et al., “Health, Life Expectancy, and Health
Care Spending.”
9. The technical appendix is available online at content.healthaffairs.org/cgi/content/full/hlthaff.w5.r18/DC2.
10. See, for example, B. Alemayehu and K.E. Warner, “The
Lifetime Distribution of Health Care Costs,” Health Services Research
39, no. 3 (2004): 627–642; and A.M. Garber, T.E. MaCurdy, and M.B. McClellan,
“Medical Care at the End of Life: Diseases, Treatment Patterns, and Costs,”
NBER Working Paper no. 6748 (Cambridge, Mass.: NBER, October 1998).
11. Garber et al. “Persistence of Medicare Expenditures.”
12. Lubitz et al., “Health, Life Expectancy, and Health
Care Spending.”
13. Alemayehu and Warner, “The Lifetime Distribution”;
and J. Lubitz, J. Beebe, and C. Baker, “Longevity and Medicare Expenditures,”
New England Journal of Medicine 332, no. 15 (1995): 999–1003.
14. D. Lakdawalla, J. Bhattacharya, and D.P. Goldman, “Are
the Young Becoming More Disabled?” Health Affairs 23, no. 1 (2004):
168–176.
15. U.S. Centers for Disease Control and Prevention, The
Burden of Chronic Diseases and Their Risk Factors (Atlanta: CDC, 2004).
Access
the table of contents for this package
Geoffrey Joyce (gjoyce{at}rand.org)
is a senior economist at RAND in Santa Monica, California. Emmett Keeler is
a senior mathematician at RAND; Baoping Shang, a fellow at the Pardee RAND Graduate
School; and Dana Goldman, corporate chair and director of health economics at
RAND.
DOI: 10.1377/hlthaff.W5.R18
©2005 Project HOPE–The People-to-People Health
Foundation, Inc.
|
