Costs of Illness for Six Major Health Conditions Among Older Adults
George Van Houtven, Ph.D.1
Amanda A. Honeycutt, Ph.D.1
Boyd Gilman, Ph.D.1
Nancy T. McCall, ScD.1
Wanda W. Throneburg, B.S.1
Kathy E. Sykes, M.A.2
November 30, 2004
1RTI International. (RTI International is a trade name of Research Triangle Institute.)
2U.S. Environmental Protection Agency, Aging Initiative
1
-------
Costs of Illness for Six Major Health Conditions Among Older Adults
ABSTRACT
Objectives. This study estimates the annual cost burden associated with six major illnesses
among the aged 65 or older population in the United States: chronic lung disease, ischemic heart
disease, stroke, lung cancer, pneumonia, and gastrointestinal illness. These illnesses were
selected because of their relatively high impact among older populations and because they
include environmental exposures as a significant risk factor.
Methods. A prevalence-based cost-of-illness approach was applied. Medical costs were
estimated from Medicare claims data, and drugs costs from Medical Expenditures Panel Survey
data. Indirect costs were estimated using a human capital approach. Morbidity-related
productivity effects were estimated through regression analyses of National Health Interview
Survey data, and mortality rates were based on National Vital Statistics Reports. All
productivity losses were valued using dollar estimates of average age-specific labor earnings and
household production.
Results. Total estimated aggregate costs range from $0.5 billion (gastrointestinal illness) to
almost $60 billion (ischemic heart disease). The combined costs of these conditions among the
aged 65 or older population in 2000 were almost $135 billion.
Discussion. With the expectation that costs of these illnesses will increase significantly as
population ages, priority should be given to prevention strategies, such as improvements in
environmental quality.
2
-------
INTRODUCTION
Since the early 1990s, health care spending in the United States has increased
dramatically, even though the percentage of the population aged 65 or older has remained
relatively stable. Between 1990 and 2002, per capita expenditures on health care grew by 45%
in real terms, while percent of population aged 65 or older remained at roughly 12.5%.
However, the aged 65 or older population is expected to increase significantly over the next
decades, and this demographic shift is certain to put additional upward pressure on aggregate
health costs and on the growing demand for health care services. To promote a better
understanding of the health-related economic implications of an aging population, this analysis
estimates the total annual cost burden associated with six major illnesses among the aged 65 or
older population in the United States.
The six categories of illness selected for this analysis are chronic lung disease (CLD),
ischemic heart disease (IHD), stroke, lung cancer, pneumonia, and gastrointestinal illness (GI).
These conditions were selected for several reasons. First, each has a relatively high rate of
prevalence among older populations (compared with younger populations) and is a major cause
of death (except GI) among persons aged 65 or older. Therefore, they are major contributors to
the total costs of illness in this age group. Second, the costs of these illnesses have rarely, if
ever, been specifically estimated for individuals aged 65 or older. Applying a consistent data set
and methods across the six illness categories not only fills a gap in the literature, it also provides
estimates that are directly comparable across the illness categories. Finally, these conditions
were selected because many cases of these illnesses are potentially preventable through
improvements in environmental conditions. All of the illnesses are known or strongly suspected
to include environmental exposures as a significant risk factor.
3
-------
Because these illnesses are often associated with exposure to environmental conditions, a
specific objective of this analysis is to inform such efforts as the U.S. Environmental Protection
Agency's (EPA) Aging Initiative, which was established in 2002 to study environmental health
impacts among older persons and to help prioritize policies for mitigating these impacts (EPA,
2003). Although the burden of disease estimates developed in this analysis are not exclusively
associated with environmental exposures, they do provide starting points (i.e., upper bounds) for
evaluating the potentially avoidable costs of illness in the aged 65 or older population that would
result from improved air, water, and general environmental quality in the United States.
Overview of Selected Health Conditions
First, a brief discussion is provided of the six health conditions with regard to their
impact on older populations in the United States and their links to environmental exposures. The
available evidence on costs of illness for each condition is also summarized. Although several
studies have estimated costs for these conditions, few have specifically focused on the aged 65 or
older population; and the variety of measurement approaches used makes it difficult to compare
results across studies.
CLD, including asthma, emphysema, and other conditions classified as chronic
obstructive pulmonary disease (COPD), affects millions of older adults (Lucas, Schiller, &
Benson, 2004). Chronic lower respiratory diseases, in particular, are the fourth leading cause of
death for persons aged 65 or older (NCHS, 2002). Exposures to common air pollutants, such as
tobacco smoke, particulate matter (PM), and ozone, have been found in several studies to cause
and/or exacerbate these conditions. For a summary of the epidemiological evidence regarding
links between common air pollutants and respiratory and cardiovascular illness, see EPA (1999).
4
-------
A number of studies have estimated costs for a range of different outcomes associated
with CLD, including hospitalization costs, expenditures for all health care services, work-loss
costs, employer costs for absenteeism and employees' health care services, and informal
caregiving costs. Consequently, cost estimates also varied widely. Estimated annual costs per
person for all health care services associated with CLD ranged from approximately $1,000 to
$7,800 (Ray et al., 2000; Wilson, Devine & So, 2000). Please note that all dollar figures in this
section have been converted to 2000 dollars using the Annual Medical Care Consumer Price
Index. Aggregate annual health care expenditures associated with CLD were estimated to range
from $2.8 billion in direct costs for adults (aged 18 or older) with asthma to $23.3 billion in
direct costs for COPD (NHLBI, 1996; Wiess, Gergen, & Hodgson, 1992).
IHD, including heart attacks and angina pectoris, is among the most common conditions
suffered by individuals aged 65 or older in the United States (Lucas et al., 2004). Diseases of the
heart are also the leading cause of death among persons aged 65 or older, accounting for roughly
33% of deaths in this age group (NCHS, 2002). Several studies have also shown that exposures
to common pollutants (e.g., PM) can increase risks associated with heart disease, and exposures
to less common toxic pollutants (e.g., lead) have also been found to increase these risks
(ATSDR, 1999; EPA, 1997 [Appendix G]; EPA, 1999).
Estimated costs for heart disease also vary widely. Estimated direct medical costs
associated with heart disease ranged from $5,900 to $7,400 per person (Druss et al., 2001; Ray et
al., 2000). Total estimated medical costs associated with heart disease ranged from $67 billion
for females aged 45 or older to more than $100 billion for all circulatory diseases in persons aged
65 or older (Hoerger et al., 1999; Hodgson & Cohen, 1999). Large differences in estimated costs
5
-------
arise primarily because of differences in the cost categories analyzed and in the age groups
considered.
Stroke affects 2 to 3 million individuals aged 65 or older each year in the United States
(Lucas et al., 2004). Cerebrovascular diseases, which include stroke, are the third leading cause
of death in the United States among persons aged 65 or older (NCHS, 2002). Several studies
have shown that exposures to pollutants, such as PM, ozone, and lead, increases risks of stroke-
related deaths (ATSDR, 1999).
A few studies estimated the average costs related to hospitalization due to stroke in older
populations. One study estimated average hospitalization costs for stroke for persons over the
age of 65 that ranged from $13,400 to almost $63,000, depending on the type of stroke
(Holloway, Witter, Lawton, Lipscomb & Samsa, 1996). Another study estimated hospital
charges of approximately $13,000 per person (aged 40 or older) for stroke alone (Mushinski,
1997). Other studies estimated total direct and indirect costs of stroke for persons of all ages to
be approximately $46 to $47 billion per year (AHA, 2004; NCHS and NHLBI, 2000).
Lung cancers are the least prevalent of the six health conditions analyzed in this report;
nevertheless, they have the second highest prevalence of cancers among persons aged 65 or older
(Lucas et al., 2004). Moreover, these cancers are known to impose significant costs on a per-
case basis. Links between environmental exposures and lung cancers are well established, with
known or suspected environmental causes including exposures to PM (Pope et al., 2002),
asbestos (ATSDR, 2001), and radon (ATSDR, 1990).
The estimated costs of lung cancers in the literature range from $8.2 billion to $32.5
billion per year using the broader definition of respiratory cancer (Brown & Fintor, 1995;
Sullivan, Ramsey, & Lee, 2000). One study estimated average Medicare payments per year for
6
-------
several cancers, including lung cancer, to be about $28,000 per affected beneficiary (Riley,
Potosky, Lubitz & Kessler, 1995).
Pneumonia is a relatively high-prevalence condition among older individuals; but more
importantly, when older individuals are afflicted with this condition, they generally require more
costly medical care than younger individuals. Pneumonia and influenza combined are the fifth
leading cause of death among persons aged 65 or older (NCHS, 2002). As with CLD, exposures
to common air pollutants, such as PM and ozone, have been found in several studies to cause
and/or exacerbate pneumonia (EPA, 1999).
Pneumonia cost studies have primarily focused on hospitalization costs. Mean or median
costs per hospitalization associated with pneumonia ranged from about $5,900 to $10,400 per
person (Fine et al., 2000; Warren et al., 2003). Variations in these estimates reflect differences
in the definitions used for pneumonia and the methods used (i.e., primary diagnosis costs versus
attributable costs). Total direct medical cost estimates associated with pneumonia ranged from
$5.7 (for persons aged 65 or older) to $19.4 billion per year (Niederman, McCombs, Unger,
Kumar & Popovian, 1998; Thom, n.d.).
GI is common in all age groups in the United States; however, persons aged 65 or older
face particularly high risks of hospitalization and death. For example, from 1979 to 1995, GI
hospitalization rates for persons older than age 65 were more than twice as high as for younger
persons. Exposures to waterborne pathogens are suspected to be significant contributors to
overall rates of GI (Bennett, Holmberg, Rogers & Solomon 1987; Morris & Levin, 1995;
Payment et al., 1991), although the attributable fraction is uncertain.
One study of Gl-related costs used a disease classification approach similar to the one
used in this analysis, which was specified to capture GI conditions most closely (although not
7
-------
exclusively) associated with environmental exposures. Sandler et al. (2002) estimated direct
medical costs associated with GI to be $1.7 billion per year for all age groups and additional
indirect costs of $540 million per year. Just over half of these costs were estimated to result
from foodborne illnesses.
METHODS
To assess the economic burden of the selected illnesses among older adults, we applied a
cost-of-illness (COI) approach that measures both the direct and indirect costs of illness. Direct
costs represent the dollar value of goods and services consumed as a result of illness and for
which payment is made. These costs include payments for treatment, diagnosis, continuing care,
rehabilitation, and terminal care. They are typically measured as costs related to hospital stays,
physician services, nursing homes, prescription drugs, and in-home health care services. Indirect
costs represent costs for which no payment changes hands but for which an economic effect is
nonetheless observed. These costs include primarily productivity losses associated with illness
and premature death; and they are typically measured as the value of lost productivity (labor and
household) due to illness.
We applied a prevalence-based approach to estimate direct and indirect costs.
Prevalence-based cost estimates include costs related to a condition for the prevalent population
over a given period (usually a year). These estimates include costs for newly diagnosed cases as
well as costs for persons in the later stages of disease. We provide estimated costs associated
with each of the six selected health conditions among individuals aged 65 or older in the year
2000.
Below we describe the data sources and methods used to estimate the different
components of direct and indirect costs. For each component, we applied a consistent set of data
8
-------
and methods across the six health conditions. We defined each health condition in a consistent
manner across data sources by using the same set of ICD-9 codes to identify service utilization
and other costs for each condition: Chronic lung disease (491-4, 496), ischemic heart disease
(410-4), stroke (430-4, 426), lung cancer (162.2-162.9, 197, 231), pneumonia (480-7), and
gastrointestinal illness (001-009 [except 008.45], 558.9).
Direct Cost Estimation
To make best use of the available national data, we separated direct costs into two
mutually exclusive components: medical costs and costs associated with self-administered
prescription drugs. A third component of direct costs that is not fully captured by this analysis is
the cost associated with nursing home care. We explored this issue by analyzing nursing home
charges using the 1999 National Nursing Home Survey (NNHS). However, findings from
NNHS indicated that none of the individual conditions had a meaningful or statistically
significant impact on annual nursing home costs. Because of these findings, we did not include
nursing home costs in the aggregate COI estimates.
Medical Costs
For individuals aged 65 or older, Medicare claims provide the most comprehensive
source of data on costs for medical services. To access a nationally representative sample of
these claims for 2000, we used data from the Consumer Assessment of Health Plans Surveys
(CAHPS) for Medicare Fee-for-Service (MFFS) beneficiaries. The CAHPS-MFFS sample
included 145,875 noninstitutionalized Medicare beneficiaries aged 65 or older. From this
sample, we analyzed Medicare claims for five main categories of services: inpatient services,
physician visits, outpatient services, home health care, and durable medical equipment. For each
of the six health conditions and five medical service categories, claims were selected based on
9
-------
their reported diagnosis codes. For inpatient services, we only selected claims recorded with a
primary diagnosis matching one of the ICD-9 codes for the condition. For outpatient and home
health services, for which Medicare claims do not report a primary diagnosis, we selected claims
recorded with any diagnosis code matching the ICD-9 codes. For professional fees and durable
medical equipment, there is a unique code for each claim.
Using these selected claims data, we estimated both the prevalence rate (number of cases
per thousand individuals) and average cost ($/case) of medical services for each condition. To
estimate the total number of cases among the entire noninstitutionalized population of
individuals aged 65 or older in 2000, we assumed that the estimated prevalence rates from
CAHPS-MFFS were applicable to the entire population of approximately 33.5 million persons.
To estimate the aggregate medical costs associated with these cases in 2000, we multiplied their
number by the corresponding average medical cost estimates for each condition.
Costs for Self-Administered Drugs
Self-administered prescription drugs are not reimbursable by Medicare and are therefore
not included in the claims-based cost estimates. To estimate these prescription drug costs for
each of the six health conditions, we used data from the 2000 Medical Expenditures Panel
Survey (MEPS), a nationally representative subsample of the National Health Interview Survey
(NHIS). The 2000 MEPS requested information about the number of prescription medication
purchases (including refills) and total expenditures for prescription drugs by sources of payment
and by condition. We first calculated total prescription drug expenditures by condition for each
individual aged 65 or older in the sample. We then calculated average prescription drug
expenditures across all individuals with a given condition. Combining these average costs with
10
-------
the prevalence estimates from the Medicare claims analysis, we estimated aggregate drug costs
for each condition.
Indirect Cost Estimation
To estimate indirect costs for each condition, we applied a human-capital approach. This
approach, commonly used in COI studies, measures the monetary value of time lost from work
or household production activities because of excess morbidity or premature mortality. These
productivity losses are estimated based on market earnings and an imputed value for household
production.
Although the human-capital approach measures key components of the economic burden
of illness on society, it has inherent limitations. For several reasons, indirect cost estimates
based on this approach are best interpreted as lower-bound estimates for the full indirect costs of
each condition. First, they do not include the value of time lost from consumption, leisure, or
volunteer activities, all of which may be higher among older individuals. Second, they do not
include the productivity or other losses incurred by informal caregivers. Third, they do not
capture the costs of pain and suffering associated with an illness.
Furthermore, although the human-capital approach measures productivity losses due to
premature mortality, these estimates should not be interpreted as measures of the full value of
lost life. Rather, they estimate one component of the economic burden resulting from premature
mortality. Primarily because of lower remaining life expectancy and labor force participation
rates among older adults, these estimates tend to be lower for older adults. These results should
not be interpreted to mean that the lives of older individuals are valued less by society.
11
-------
Morbidity
Using data from the 2001 NHIS, we estimated the impact of specific health conditions on
labor and household productivity among persons aged 65 or older. For labor productivity losses,
we estimated the losses associated with being completely unable to work (out of the labor force)
due to the health condition. These estimates were based respondents who reported that they were
not currently working but did work previously. We estimated a probit model of not working for
those who reported working previously and included controls for age and for three of the health
conditions for which adequate data were reported: CLD, IHD, and stroke. The resulting probit
model coefficients were used to calculate the marginal effect of having each of the three health
conditions on the probability of not currently working for persons who reported working
previously. The marginal effects are the estimates of the impact of each health condition on the
probability of being unable to work because of the condition.
To estimate household productivity losses, we used a regression approach to analyze the
determinants of the number of reported bed days by NHIS respondents aged 65 or older. By
controlling for several other factors, including education, poverty status, self-reported health
status, labor force participation status, and smoking status, the regression analysis allowed us to
estimate the incremental effect of specific conditions on number of bed days. Based on these
results, we predicted the average number of bed days attributable to each condition.
To estimate the dollar values associated with the changes in labor force participation and
bed days attributable to each condition, we applied existing estimates of age-specific annual
earnings and household productivity values for 2000. Using values reported in Grosse (2003),
we calculated average annual labor force earnings of $31,052 for individuals aged 65 or older
who are in the labor force with positive earnings. The average annual value of household
12
-------
services produced by persons aged 65 or older was estimated to be $12,936, an average of $35
per day.
Mortality
To assess mortality-related productivity losses, we used data from the 1998 National
Vital Statistics Report (NVSR), which provides statistics on death rates in 1996 by age group and
cause of death. The 1998 NSVR was the most recent NSVR that used ICD-9 codes to categorize
causes of death and could therefore be most closely matched with the illness classification
approach used in our analysis. We applied these rates to determine the number of annual deaths
(by age groups) that are primarily associated with each of the six health conditions.
To estimate the dollar value of productivity losses associated with premature mortality,
we multiplied the cause-specific and age-group-specific number of deaths for each health
condition by estimates of age-group-specific present value of earnings (earnings and household
production), which were also estimated by Grosse (2003). Because labor and household
production estimates for the aged 85 or older group were not available for this analysis, we were
not able to estimate mortality-related productivity losses for this group. Average productivity
losses for persons aged 85 or older are expected to be relatively low compared with younger age
groups; however, by not including values for this age group, we have underestimated total
mortality-related losses for each condition.
RESULTS
Using the data and methods described above, we estimated the costs of illness associated
with each of the six selected health conditions in 2000 among individuals aged 65 or older.
These results are summarized below for each condition (in 2000 dollars). We first summarize
the estimates of (1) rates of illness and mortality, and (2) average direct and indirect costs of
13
-------
illness for each condition. We then combine these values and report the estimates of aggregate
costs.
Prevalence and Average Direct Costs
Table 1 summarizes the estimates of the prevalence and the direct medical costs
associated with the six health conditions based on the CAHPS-MFFS beneficiary data. The
percent of beneficiaries aged 65 or older with each condition ranged from 1% for lung cancer to
almost 18% for IHD. CLD and stroke had the second and third highest prevalence rates
affecting, respectively, 11% and 7% of this population.
[Table 1 about here]
Although the prevalence of lung cancer was low among the aged 65 or older Medicare
population, per capita medical costs were the highest of the six health conditions, at more than
$7,500 per case per year. This annual cost was over twice the estimated amount for the next
highest cost illnesses—pneumonia, IHD, and stroke—which imposed direct medical costs of
roughly $3,000 per case per year. Also, GI was associated with the lowest average costs, at less
than $600 per case per year.
Although not reported in Table 1, the analysis of CAHPS-MFFS data also found the
following for all six health conditions:
• Inpatient services accounted for the largest share of average medical costs (compared
with physician, outpatient, home health, and durable medical equipment costs),
ranging from 37% for CLD to 83% for pneumonia.
• Rates of illness were higher among persons aged 75 or older compared with persons
aged 65 to 74. This difference was most notable for stroke and pneumonia, but was
not statistically significant (at a 0.05 level) for lung cancer.
14
-------
• Rates of illness were higher for males aged 65 or older than for females aged 65 or
older (with the exception of GI, which had similar rates for males and females in this
age group).
Using MEPS data as described above, we also estimated the average cost of self-
administered prescription drugs for each condition. The number of MEPS respondents aged 65
or older who reported prescription medication purchases for lung cancer was too small to reliably
include in this analysis; however, the numbers for the other conditions ranged from 114
respondents for stroke to 174 respondents for CLD. Based on these observations, we estimated
average self-administered prescription drug costs ranging from $35 per case per year for
pneumonia and GI to $363 per case per year for CLD.
Average Productivity Losses Due to Morbidity
Using data from the 2001 NHIS and from Grosse (2003), we estimated average
morbidity-related labor productivity losses for three of the six health conditions. These results
are presented in Table 2. Pneumonia and GI were not included in this part of the analysis
because NHIS did contain information about these illnesses. Of the 6,139 respondents aged 65
or older in the NHIS sample, approximately 8% indicated that they were unable to work but did
work previously. Estimates of the impact of having each health condition on the probability of
being unable to work, controlling for age and the other conditions, were 2% for IHD, 4% for
CLD, and 6% for stroke. These estimated probabilities are presented in Table 2.
[Table 2 about here]
As shown in Table 2, the average earnings for individuals aged 65 or older—conditional
on being in the labor force—is approximately $31,000 per year. Multiplying the illness-specific
incremental probability of being unable to work by the average conditional earnings provides an
15
-------
estimate of the average productivity loss due to inability to work for individuals aged 65 or older
who have the illness in question. These average annual labor productivity losses were estimated
to range from $621 per person with IHD to almost $1,900 per person with stroke.
Using NHIS data, we also estimated household productivity losses for three of the six
health conditions. We restricted the sample to all persons aged 65 or older, regardless of
employment status, which resulted in a sample of 3,950 older adults. Because of the discrete
nature of the variable of interest for this analysis (i.e., number of days spent in bed during the
year), we used a negative binomial regression with a log link to estimate the determinants of bed
days. The regression results are presented in Table 3. Controlling for sociodemographic
characteristics—age, gender, education, income, work status, and race—and smoking status, the
results indicate that CLD, heart disease, and stroke all have positive effects on the number of bed
days experienced during the year. The effects of stroke and heart disease were found to be
significant at a level < 0.01, whereas CLD was significant at a 0.06 level.
[Table 3 about here]
Based on the regression results, we estimated the average number of bed days attributable
to each health condition for individuals aged 65 or older. For each individual in the analysis
sample with one of the conditions (525 with stroke, 749 with CLD, and 1,891 with IHD), we first
used the regression results to estimate the difference in predicted bed days with and without the
condition (i.e., calculating predicted bed days for condition=0 as compared with condition=l).
We then estimated the average of these differences for each health condition subsample. The
average number of bed days per year associated with each illness varied from 5 days per case of
CLD to 8 days per case of IHD to 14 days per case of stroke. To estimate the indirect costs
associated with each bed day, we assumed that they each resulted in one day of lost household
16
-------
productivity. Based on estimates from Grosse (2003), the average value of household
productivity for individuals aged 65 or older was equivalent to approximately $35 per day in
2000. This estimate is based on the average time spent in household production (e.g.,
housework, yard maintenance, etc.), care providing, and personal care activities, each valued at
the average market wage rate. Multiplying the average number of attributable bed days for each
illness by this value produced estimates of average annual household productivity losses equal to
$185 per case of CLD, $278 per case of IHD, and $512 per case of stroke.
Average Productivity Losses Due to Mortality
To estimate the number of deaths in 2000 associated with each category of illness among
individuals aged 65 or older, we applied age- and illness-specific mortality rates for 1996. In the
1998 NSVR, these rates were reported for three age groups of interest: 65 to 74, 75 to 84, and 85
or older. Assuming that the percentage of the population in each age group who died due to each
illness did not change significantly from 1996 and 2000, we applied these percentages to the
2000 U.S. population in each age group to estimate the total number of deaths associated with
each illness. These estimates are presented in Table 4. Of the six health conditions, IHD is
responsible for the highest number of deaths in all three age categories, with an estimated total of
almost 450,000 deaths among persons aged 65 or older. GI deaths account for by far the fewest
number of estimated deaths (less than 1,000). Total deaths associated with each of the other four
health conditions were estimated to fall between 80,000 and 160,000 in 2000.
[Table 4 about here]
To estimate productivity losses associated with these deaths, we again used results from
Grosse (2003); however, in this case we used the average present value of future earnings and
household production (assuming a 5 percent discount rate) for each age group. As shown in
17
-------
Table 4, the present value of labor and household productivity that is lost due to premature death
in 2000 is approximately $190,000 for individuals aged 65 to 74 and $94,000 for individuals
aged 75 to 84. Unfortunately, present value estimates were not available for individuals aged 85
or older; therefore, total mortality-related productivity losses in this age group could not be
calculated.
Aggregate Direct and Indirect Costs of Illness
Combining the estimated rates of morbidity and mortality with the average direct and
indirect costs reported above, we estimated aggregate costs of illness for each of the six health
conditions (see Table 5). First, to estimate the total number of cases of each illness, we applied
the estimated prevalence rate from the Medicare beneficiary population aged 65 or older (see
Table 1) to the entire aged 65 or older population in 2000 (33.5 million). The total number of
deaths attributable to each illness was calculated by summing across the three age categories
reported in Table 4. IHD accounts for both highest number of illnesses (over 6 million) and
deaths. Lung cancers account for the fewest number of illnesses (355,000), but more deaths than
CLD, pneumonia, or GI.
[Table 5 about here]
Aggregate direct costs for IHD were found to be significantly higher than for the other
conditions. Cases of IHD among persons aged 65 or older were estimated to impose medical and
drug costs of almost $20 billion per year in 2000. By comparison, CLD, stroke, and pneumonia
had direct cost estimates of between $5 billion and $10 billion. Due primarily to its low rate of
prevalence, aggregate direct costs associated with lung cancer were estimated to be less than $3
billion (although this estimate does not include drug costs), and for GI they were less than $0.5
billion.
18
-------
Aggregate indirect costs resulting from morbidity, which were estimated for three of the
six health conditions (CHD, IHD, and stroke), were found in each case to be between $5-6
billion per year. Due to data limitations, productivity losses could not be estimated for the other
three conditions, which are less prevalent in the aged 65 or older population. Aggregate indirect
costs associated with premature mortality were found to be highest for IHD (almost $35 billion),
followed by lung cancer (approximately $15 billion), and CLD and stroke (both close to $10
billion). Due to data limitations, none of these estimates includes mortality-related productivity
losses among persons aged 85 or older. The resulting underestimation of productivity losses is
largest in absolute terms for IHD because over 170,000 deaths are attributed to IHD among the
aged 85 or older population (see Table 4) and largest in relative terms for pneumonia because
more than 50% of pneumonia-caused deaths are from this age group.
Summing the direct and indirect cost estimates for each condition, the total estimated
aggregate costs range from less than $1 billion for GI to almost $60 billion for IHD. The
combined cost of these conditions among the aged 65 or older population in 2000 was estimated
to be almost $135 billion.
DISCUSSION
The population of older adults in the United States is expected to increase rapidly over
the next 50 years. With this trend will come rising burdens of illness and increasing demands on
the public health system. In this analysis, we examined the current aggregate cost burden
imposed by six major health conditions among older adults, and estimated total direct and
indirect costs of $135 billion for 2001. With the aged 65 or older population expected to more
than double by 2030, it would not be surprising to see these costs increase by a similar
proportion.
19
-------
Despite the magnitude of these estimated costs, it is important to emphasize that for
several reasons they are best interpreted as lower-bound estimates of the total annual societal
losses associated with the illnesses studied. First, a well-recognized limitation of the COI
approach is that it cannot capture all of the relevant losses associated with illness, and for
illnesses affecting older adults, this limitation is particularly important. The COI approach does
not measure the (negative) value of pain and suffering for those directly or indirectly affected by
the condition, nor does it capture the value of lost leisure time. Moreover, in cases where the
conditions lead to premature mortality, the COI approach cannot measure the full value to
society of avoiding those deaths. What the COI method does is measure the portion of economic
losses that are most directly observable from available data on medical expenditures, labor
market participation and earnings, and household production activities. These measurable losses
represent a true burden on society, but they do not tell the entire story. For older adults who, on
average, spend relatively less time in paid-labor activities and more time in leisure-related
activities, unmeasured losses are likely to be particularly large.
Second, due to data limitations, this analysis does not include certain direct and indirect
costs, such as nursing home costs, reductions in work days (as opposed to leaving the
workforce), and mortality-related productivity losses for individuals aged 85 or older. These
omitted costs are expected to be relatively small compared with the measured costs, but
unfortunately they could not be reliably measured.
Third, estimating condition-specific medical costs based on Medicare claims data is
complicated by the fact that inpatient services are often reported with more than one diagnosis.
To estimate inpatient costs, we used a conservative approach that is likely to underestimate these
costs. We only included costs if the condition of interest was listed as the primary diagnosis.
20
-------
This approach includes some costs that may be associated with comorbid conditions, but it also
omits a portion of the costs attributable to the condition of interest if the condition is secondary
to the first reported diagnosis.
Given the magnitude of the current cost burden associated with these illnesses and the
expectation that these costs will increase significantly as the U.S. population ages, it is
particularly important to consider strategies for preventing these illnesses. Key among these
strategies are policies to improve environmental health (one of the top 10 focus areas identified
in Healthy People 2010). All of the six health conditions are known to be associated, at least in
part, with environmental exposures, and older adults are often more susceptible than younger
adults to environmental exposures. Therefore, improving air and water quality can play an
important role in preventing and limiting the cost burden associated with these conditions in the
aged 65 or older population. Future research should focus on understanding what proportion of
the estimated costs of illness are specifically attributable to environmental exposures and to other
preventable causes of illness.
21
-------
ACKNOWLEDGMENTS
This study was funded by the U.S. Environmental Protection Agency's Aging Initiative
through a subcontract with ICF Consulting, Inc. (Prime contract #: 68-W-02-045). We would
like to thank Bill O'Neil for his helpful comments and suggestions.
22
-------
REFERENCES
Agency for Toxic Substance Disease Registry (ATSDR). (1990). Toxicological profile for
radon. PB/91/180422/AS.
Agency for Toxic Substance Disease Registry (ATSDR). (1999). Toxicological profile for lead.
PB/99/166704.
Agency for Toxic Substance Disease Registry (ATSDR). (2001). Toxicological profile for
asbestos. PB/2001/109101.
American Heart Association. Heart Disease and Stroke Statistics—2004 Update. Economic cost
of cardiovascular diseases, from
.
Bennett, J. V., Holmberg, S. D., Rogers, M. F., & Solomon, S. L. (1987). Infections and
parasitic diseases. American Journal of Preventive Medicine, 3, 102-114.
Brown, M. L., & Fintor, L. (1995). The economic burden of cancer. In P. Greenwald, B. S.
Kramer, & D. L. Weed (Eds.), Cancer prevention and control. New York: Marcel
Dekker (pp. 69-81. As cited in Kirschstein, R. (2000). Disease-specific estimates of
direct and indirect costs of illness and NIH support. Washington, DC: Office of the
Director, National Institutes of Health, Department of Health and Human Services.
Druss, B. G., Marcus, S. C., Olfson, M., Tanielian, T., Elinson, L., & Pincus, H. A. (2001).
Comparing the national economic burden of five chronic conditions. Health Affairs
(Millwood), 20(6), 233-241.
23
-------
Fine, M. J., Pratt, H. M., Obrosky, D. S., Lave, J. R., Mcintosh, L. J., Singer, D. E., Coley, C.
M., & Kapoor, W. N. (2000). Relation between length of hospital stay and costs of care
for patients with community-acquired pneumonia. American Journal of Medicine,
109(5), 378-385.
Grosse, S. D. (2003). Productivity loss tables (Appendix). In A. C. Haddix, S. M. Teutsch, P.,
& S. Corso, (Eds.), Prevention effectiveness: A guide to decision analysis and economic
evaluation. 2nd ed. London: Oxford University Press.
Hodgson, T. A., & Cohen, A. J. (1999). Medical expenditures for major diseases, 1995. Health
Care Financing Review, 21(2), 119-164.
Hoerger, T. J., Downs, K. E., Lakshmanan, M. C., Lindrooth, R. C., Plouffe, L., Wendling, B.,
West, S. L., & Oshfeldt, R. L. (1999). Health care use among U.S. women aged 45 and
older: Total costs and costs for selected postmenopausal health risks. Journal of
Women's Health and Gender-Based Medicine, 5(8), 1077-1089.
Holloway, R. G., Witter Jr., D. M., Lawton, K. B., Lipscomb, J., & Samsa, G. (1996). Inpatient
costs of specific cerebrovascular events at five academic medical centers. Neurology,
46(3), 854-860.
Lucas, J. W., Schiller, J. S., & Benson, V. (2004). Summary health statistics for U.S. adults:
National Health Interview Survey, 2001. National Center for Health Statistics. Vital
and Health Statistics 70(218).
Morris, R. D., & Levin, R. (1995). Estimating the incidence of waterborne infectious disease
related to drinking water in the United States. In Assessing and managing health risks
from drinking water contamination: Approaches and applications. Proceedings of the
Rome Symposium, September 1994.
24
-------
Mushinski, M. (1997). Variations in average charges for strokes and TIAs: United States,
1995. Statistical Bulletin-Metropolitan Life Insurance Company, 75(4), 9-18.
National Center for Health Statistics (NCHS), Centers for Disease Control and Prevention
(Thomas A. Hodgson) and National Heart, Lung, and Blood Institute (NHLBI), NIH
(Thomas J. Thorn). As cited in Kirschstein, R. (2000). Disease-specific estimates of
direct and indirect costs of illness and NIH support. Washington, DC: Office of the
Director, National Institutes of Health, Department of Health and Human Services.
National Center for Health Statistics (NCHS), Centers for Disease Control and Prevention.
(2002). Deaths, percent of total deaths, and death rates for the 15 leading causes of death
in selected age groups, by race and sex: United States, 1999-2001. Retrieved May 18,
2004, from http://www.cdc.gov/nchs/data/dvs/LCWK3_2000.pdf
National Heart, Lung, and Blood Institute (NHLBI). (1996). Morbidity and mortality chartbook
on cardiovascular, lung, and blood diseases. As cited in Kirschstein, R. (2000).
Disease-specific estimates of direct and indirect costs of illness and NIH support.
Washington, DC: Office of the Director. National Institutes of Health, Department of
Health and Human Services.
Niederman, M. S., McCombs, J. S., Unger, A. N., Kumar, A., & Popovian, R. (1998). The cost
of treating community-acquired pneumonia. Clinical Therapeutics, 20(4):820-837.
Payment, P., Richardson, L., Siemiatycki, J., Dewar, R., Edwardes, M., & Franco, E. (1991). A
randomized trial to evaluate the risk of gastrointestinal disease due to consumption of
drinking water meeting current microbiological standards. American Journal of Public
Health, 57(6), 703-708.
25
-------
Pope, C. A., Burnett, R. T., Thun, M. J,. Calle, E. E., Krewski, D., Ito, K., & Thurston, G. D.
(2002). Lung cancer, cardiopulmonary mortality, and long-term exposure to fine
particulate air pollution. Journal of the American Medical Association, 287(9),
1132-1141.
Ray, G. T., Collin, F., Lieu, T., Fireman, B., Colby, C. J., Quesenberry, C. P., Van den Eeden, S.
K., & Selby, J. V. (2000). The cost of health conditions in a health maintenance
organization. Medical Care Research and Review, 57(1), 92-109.
Riley, G. F., Potosky, A. L., Lubitz, J. D., & Kessler, L. G. (1995). Medicare payments from
diagnosis to death for elderly cancer patients by stage at diagnosis. Medical Care, 33(8),
828-841.
Sandler, R. S., Everhart, J. E., Donowitz, M., Adams, E., Cronin, K., Goodman, C., Gemmen, E.,
Shah, S., Avdic, A., & Rubin, R. (2002). The burden of selected digestive diseases in
the United States. Gastroenterology, 722(5), 1500-1511.
Sullivan, S. D., Ramsey, S. D., & Lee, T. A. (2000). The economic burden of COPD. Chest,
777(2 Suppl), 5S-9S.
Thom, T. (n.d.). Unpublished estimates. As cited in Kirschstein, R. (2000). Disease-specific
estimates of direct and indirect costs of illness and NIH support. Washington, DC:
Office of the Director, National Institutes of Health, Department of Health and Human
Services.
U.S. Environmental Protection Agency (EPA). (1997). The benefits and costs of the Clean Air
Act, 1970 to 1990. EPA 410-R-97-002. Washington, DC: Office of Air and Radiation.
U.S. Environmental Protection Agency (EPA). (1999). The benefits and costs of the Clean Air
Act, 1990 to 2010. EPA410-R-99. Washington, DC: Office of Air and Radiation.
26
-------
U.S. Environmental Protection Agency. (2003). Aging initiative: A national agenda for the
environment and the aging. Retrieved December 21, 2003, from
http://www.epa.gov/aging/pdfs/factsheet.pdf
Warren, D. K., Shukla, S. J., Olsen, M. A., Kollef, M. H., Hollenbeak, C. S., Cox, M. J., Cohen,
M. M., & Fraser, V. J. (2003). Outcome and attributable cost of ventilator-associated
pneumonia among intensive care unit patients in a suburban medical center. Critical
Care Medicine, 31(5), 1312-1317.
Wiess, K. B., Gergen, P. J., & Hodgson, T. A. (1992). An economic evaluation of asthma in the
United States. New England Journal of Medicine, 356, 862-866.
Wilson, L., Devine, E. B., & So, K. (2000). Direct medical costs of chronic obstructive
pulmonary disease: Chronic bronchitis and emphysema. Respiratory Medicine, 94,
204-213.
27
-------
Table 1. Prevalence and Average Direct Medical Costs for Six Health Conditions: Annual
Medical Payments in 2000 for Medicare Beneficiaries Aged 65 or Older
Direct Medical Costs
Beneficiaries with a Diagnosis ($/Case/Year)
Health Condition
Number
Percent
CI (±)
Mean
CI(±)
Chronic Lung Disease
2,871,630
11
(10.8, 11.1)
$1,830
(1,748, 1,913)
Ischemic Heart Disease
4,696,071
18
(17.7, 18.1)
$3,003
(2,880, 3,125)
Stroke
1,924,797
7
(7.2, 7.5)
$2,812
(2,653, 2,971)
Lung Cancer
276,900
1
(1.0, 1.1)
$7,751
(6,697, 8,534)
Pneumonia
1,288,106
5
(4.8, 5.0)
$3,052
(2,886, 3,218)
Gastrointestinal Illness
588,371
2
(2.2, 2.3)
$579
(522, 634)
Note: Geographic weights used to obtain unbiased nationally representatives estimates for Medicare 65+
population.
28
-------
Table 2. Average Labor Productivity Losses Due to Inability to Work, by Health
Condition
Average Annual
Earnings for
Estimated
Persons Aged 65 or
Average Labor
Probability of Being
Older in Labor
Productivity Loss
Health Condition
Unable to Work
Force
($/Case/Year)
Chronic Lung Disease
4%
$31,052
$1,273
Ischemic Heart Disease
2%
$31,052
$621
Stroke
6%
$31,052
$1,894
Note: Estimates are in 2000 dollars and for persons aged 65 or older.
29
-------
Table 3. Determinants of Bed Days: Negative Binomial Regression Results
Independent Variable: Number of Bed Days in 2002 (N=3,950)
Explanatory Variable
Coefficient
High school graduate
-0.16
Female
0.27b
Low income (<200% of poverty line)
0.03
Middle income (<500% of poverty line)
0.08
High income (> 500% of poverty line)
-0.52
Former smoker
0.29
Never smoked
0.01
Smoking status unknown
-25.08a
Non-Hispanic white
-0.25
Non-Hispanic black
0.21
Non-Hispanic other
0.41
Retired
0.79a
Not working but previously worked
2.21a
Never worked
1.30a
No functional limitation reported
-1.29a
Age (in years)
0.02
Self-report of CHRONIC LUNG DISEASE
0.408b
Self-report of HEART DISEASE
0.73a
Self-report of STROKE
0.81a
Constant term
-0.65
30
-------
Notes: All explanatory variables except constant term and age are indicator/dummy variables.
Reference case is Hispanic male, high school dropout, current smoker, with income below 100% of the poverty line,
who currently works, and has a self-reported functional limitation, but does not have CLD, IHD, or stroke history.
a denotes P-value based on robust standard errors < 0.05.
b denotes P-value based on robust standard errors <0.10 and >0.05.
31
-------
Table 4. Deaths and Mortality-Related Productivity Losses in 2000, by Health Condition
Average Mortality-Related Productivity
Estimated Number of Deaths by Age Group Losses by Age Group3
Health Condition
65 to 74
75 to 84
85 or older
65 to 74
75 to 84
85 or older
Chronic Lung Disease
29,869
44,651
23,806
$190,000
94,000
n/a
Ischemic Heart Disease
97,226
169,993
176,275
$190,000
94,000
n/a
Stroke
25,053
59,972
70,975
$190,000
94,000
n/a
Lung Cancer
56,402
47,156
12,802
$190,000
94,000
n/a
Pneumonia
10,408
28,789
44,067
$190,000
94,000
n/a
Gastrointestinal Illness
98
212
263
$190,000
94,000
n/a
a Average present value of future labor and household productivity,
n/a = not available.
32
-------
Table 5. Estimated Cases and Aggregate Costs of Illness for Six Health Conditions Among Individuals Aged 65 or Older in
2000
Cases (in thousands) Direct Costs ($ millions) Indirect Costs ($ millions) Total Costs
Health Condition
Illnesses
Deaths
Medical
R\ Drug
Subtotal
Morbidity
Mortality
Subtotal
($ millions)
Chronic Lung Disease
3,685.8
98.3
6,745.0
1,337.0
8,082.0
5,374.6
9,895.8
15,270.5
23,352.5
Ischemic Heart Disease
6,027.5
443.5
18,100.7
1,648.8
19,749.5
5,416.2
34,535.8
39,951.9
59,701.4
Stroke
2,470.5
156.0
6,947.1
371.8
7,318.9
5,945.4
10,423.7
16,369.1
23,688.0
Lung Cancer
355.4
116.4
2,754.8
n.e.
2,754.8
n.e.
15,182.4
15,182.4
17,937.2
Pneumonia
1,653.3
83.3
5,045.9
57.3
5,103.2
n.e.
4,695.7
4,695.7
9,798.9
Gastrointestinal Illness
755.2
0.6
437.3
26.6
463.9
n.e.
38.7
38.7
502.5
n.e. = not estimated
33
------- |