1
Measuring Nursing Home Price Growth between 2000-2009
1
Tina Highfill & David Johnson
U.S. Bureau of Economic Analysis
ABSTRACT
The proper measurement of inflation in health care is important for policymakers to
understand the drivers of price growth. For this reason, the U.S. Bureau of Economic Analysis
(BEA) recently released an alternate presentation of inflation for the health care sector that
examines prices by disease, such as treatment of diabetes, rather than by place of service, such as
a hospital stay. This account does not yet incorporate spending on nursing home care, providing
an incomplete picture of inflation in the health care sector. To fill this gap, this paper calculates
price indexes by disease for nursing home care for 2000-2009. We find prices in the overall
nursing home sector grew at an average annual rate of only 0.9% during the period using these
indexes. Price growth was slower for long-term nursing home residents (1.4%) compared to
short-term residents (2.8%). Diseases of the circulatory system was the most prevalent disease
category, followed by mental illness for long-term residents and diseases of the musculoskeletal
system and connective tissue for short-term patients. These three categories of diseases also
received the largest allocations of spending, with the bulk going towards patients diagnosed with
mental conditions. Overall, nursing home price growth in the 2000s was much slower than for
other health care sectors. Incorporating disease-based price indexes for nursing homes into
BEA’s new health care account will provide a more comprehensive picture of health care
spending trends and inflation.
1
We thank Ana Aizcorbe, Abe Dunn, and Anne Hall for useful comments. The views expressed
in this paper are solely those of the authors and not necessarily those of the U.S. Bureau of
Economic Analysis or the U.S. Department of Commerce.
2
INTRODUCTION
Spending on nursing homes represented over 5% of all national health care expenditures,
about $149 billion in 2011
2
(National Center for Health Statistics, 2014). Almost 5% of all
Medicare expenditures are spent on short-term nursing home stays (Medicare Current
Beneficiary Survey, 2012). Over 1.4 million people were long-term residents of a nursing home
in 2012 (Centers for Medicare & Medicaid Services, 2013). More than 40% of long-term care
spending is paid by Medicaid, representing almost one-third of Medicaid’s total annual
expenditures (Paradise, 2015). The demand for long term care services is only expected to
increase as the population continues to age and deal with increasingly complex conditions (U.S.
Department of Health & Human Services, 2010). Given the large scale of public spending
attributed to the nursing home sector and the number of people involved in this type of care, the
proper measurement of inflation in the nursing home care is important to understand what is
driving spending growth.
To better understand price growth in the health care sector, the U.S. Bureau of Economic
Analysis (BEA) recently released a “health care satellite account” that measures medical care
inflation using new methods recommended by the Committee on National Statistics (Dunn et al.,
2015; National Research Council, 2010). This account estimates price growth by allocating
spending to disease categories and calculating medical care expenditure (MCE) indexes. These
MCE indexes redefine output in the health care sector as the treatment of disease (e.g., cancer or
diabetes), the suggested method of health economists to better understand drivers of health care
expenditures and the returns to spending (Berndt et al., 2000). The MCE indexes differ from
2
Nursing home spending represents nearly 9% of personal consumption expenditures (PCE) on
health care services (U.S. Bureau of Economic Analysis, 2015).
3
BEA’s official method of calculating prices and spending for health care, which is measured by
place of service, such as a stay in a hospital or a doctor’s visit. An MCE index picks up shifts in
the treatment of diseases that the official index does not. For example, MCE indexes capture the
effect of substitutions between places of service that occur from changes in technology or
reimbursement, such as certain procedures shifting away from expensive hospital stays to less-
expensive outpatient ambulatory surgical centers (Rosenberg & Browne, 2001). Additionally,
MCE indexes account for changes in insurance coverage that can impact the cost of care. This is
important in the nursing home sector, where a large proportion of nursing home residents shift
from paying out-of-pocket to having Medicaid coverage, which often reimburses nursing homes
at lower rates (Wiener et al., 2013). An MCE index captures this shift as a drop in price, whereas
the official method does not. BEA’s new account covers the bulk of medical care spending in the
U.S., but does not yet include spending on care in nursing homes. Incorporating this missing
piece of spending will provide a more comprehensive picture of health care and the drivers of
price growth in the U.S. Therefore, the objective of this paper is to calculate MCE indexes for
nursing homes that can be incorporated into BEA’s health care satellite account. The components
of the MCE indexes are examined to identify which diseases account for the most spending and
growth in prices. Finally, the MCE indexes are compared to BEA’s official nursing home index,
the Personal Consumption Expenditures (PCE) index.
The rest of the paper is organized as follows: the Background section explains nursing home
pricing and discusses previous research on nursing home price growth. The Methods section
explains the data and calculations used to estimate spending by disease and the MCE indexes. It
also includes a short description of the conventional method of measuring nursing home price
growth using PCE indexes. Results are presented with a discussion of the implications for adding
4
nursing home spending to BEA’s health care satellite account and the reasons for different price
growth between the MCE and PCE indexes.
Background
Most nursing home residents are Medicare beneficiaries, but the majority of overall
nursing home spending comes from Medicaid (U.S. Department of Health & Human Services,
2010). Medicare only pays for short-term nursing home care when it occurs directly after an
eligible hospital stay, usually one that was three days or longer. Most nursing home patients are
permanent or “long-term” residents. They typically pay for nursing home care from personal
funds, or use Medicaid if they do not have the resources. Medicaid pays for the majority of total
spending on long-term nursing home care (National Center for Health Statistics, 2014).
Short-term or skilled nursing facility (SNF) stays represented about 30% of spending on
total nursing home care in 2009 (Medicare Current Beneficiary Survey, 2012). The price for a
SNF stay is determined by Medicare’s Prospective Payment System (PPS). The PPS prices
nursing homes stays based on the patient’s condition and degree of resources expected to be
utilized, similar to diagnosis-related groups used to reimburse hospital services. The first 20 days
of a SNF stay is paid entirely by Medicare, while stays that last 20-100 days require a co-pay
from the beneficiary, currently $157.50 per day (Centers for Medicare & Medicaid Services,
2015). Beneficiaries typically pay out-of-pocket or are covered by Medicaid. In 2004, Medicare
paid for almost two-thirds of all spending on SNF stays, while Medicaid paid for 16% and
private coverage paid for 11% (Figure 1.a). When a nursing home stay reaches 101 days,
Medicare no longer pays and patients either leave the nursing home or become long-term
residents. When comparing daily reimbursement rates, SNF stays are generally higher-priced
5
than for long-term nursing home care because SNF patients are usually receiving more intensive
care, such as rehabilitative services after a broken hip. While the SNF patients cost more using
average daily rates, annual spending is higher for long-term patients because the nursing home
stay covers a much longer period of time, often the entire year (see Figures 2.a and 2.b).
Nursing home services for long-term patients are broken into two categories: custodial
and ancillary. The custodial category covers room and board. The ancillary category covers all
others services, such as physical therapy and prescription medications. While nursing homes
typically charge all residents the same amount for services, Medicaid usually reimburses nursing
homes at lower rates than private payers (Zuckerman et al., 2009). In the 2000s, Medicaid paid
nursing homes a set daily rate for custodial services for each resident. The rate varies by nursing
home depending on the case mix, a value determined from the average patient severity in each
nursing home. Spending for ancillary services differs by patient, depending on their condition
and the payer. For example, a patient with multiple chronic conditions may have greater
spending than a patient of lesser severity due to higher costs for prescription medication. The
payer matters because a private patient can choose services that a Medicaid patient often cannot,
such as wanting a private room or requesting a brand name medication instead of a generic.
Stewart et al. (2009) provide comprehensive estimates of nursing home price growth in
the U.S. from 1977-2004. They look at changes in per diem prices for nursing home patients
using data from the National Nursing Home Survey (NNHS). Stewart et al. (2009) found overall
nursing home price growth slightly outpaced prices for other medical care from 1977-2004. The
authors suggest the relatively fast growth in nursing home prices reflects increased costs due to
investments in quality improvements, specifically, an increase in expensive specialty units for
patients with cognitive impairment and an increase in full-time nursing staff ratios. They also
6
found prices for out-of-pocket payers grew slightly faster than Medicaid prices. This is not
surprising since Medicaid payments are determined annually by each state, whereas prices for
private patients are not subject to the same constraints.
There is considerable market variation in average per diem nursing home prices for
patients paying out-of-pocket, ranging from $135 to over $520 (Northwestern Mutual, 2013).
Market factors have been shown to impact private nursing home prices above and beyond the
costs borne by the facility. The supply of nursing homes steadily declined in the 2000s, dropping
almost 10% during the decade (Centers for Medicare & Medicaid Services, 2013). Additionally,
substitutes to nursing home care have increased competition in the sector, such as home health
care and adult day care (Grabowski et al., 2012). Public reporting of nursing home quality
indicators has also been shown to impact nursing home prices for private payers. Clement et al.
(2012) found low-quality nursing homes in Wisconsin responded to public quality reporting by
increasing prices for private payers. A positive relationship between public reporting and prices
was also found in a study that used a national sample of nursing homes from 2001-2006
(Mukamel et al., 2010).
METHODS
Data
Nursing home payment and utilization data were taken from the Medicare Current
Beneficiary Survey (MCBS), an annual survey of Medicare beneficiaries on health care
utilization and expenses. All health care utilization and spending are collected in the MCBS, not
just medical expenses covered by Medicare. The MCBS also covers aspects such as patient
demographics and measures of health status. The MCBS sample uses a 2-year overlapping panel
7
design and has a sample size of approximately 12,000 beneficiaries per year, of which about
1,200 have at least one nursing home stay (Medicare Current Beneficiary Survey, 2012).
There are three types of nursing home patients: those with only long-term stays, those
with only short-term (SNF) stays, and those who had both in the year. Beneficiaries with both
types of stays usually represent a patient that became a long-term resident after a short-term stay
following a hospital stay. Disease information for short-term patients is available from billing
claims diagnoses and also from an annual survey. Up to nine different International
Classification of Disease codes (ICD-9) are available on each claim. These diagnosis codes are
mapped to 260 mutually exclusive clinical classification codes using the Clinical Classification
Software (CCS) from the Agency for Healthcare Research and Quality. The CCS has been used
in the past to group diseases into clinically meaningful groups (Aizcorbe & Nestoriak, 2011) and
is currently the classification system used to define diseases in BEA’s new health care satellite
account.
Information on spending and medical conditions for long-term patients in the MCBS
comes from an annual patient survey. The survey asks about the prevalence of over two dozen
major conditions. A nurse completes a questionnaire for each survey resident and indicates
which of any of the diseases the patient has. For example, the survey asks “does the patient have
dementia?” The survey diseases and conditions can be mapped to exclusive CCS codes. For a
full concordance of the survey conditions and CCS categories, see Appendix Table 1 of Hall and
Highfill (2013). While the survey does not cover all possible ICD-9 conditions, half of all the 18
disease chapters are represented. Two of the nine chapters excluded from the survey are related
to pregnancy and children. The other seven were found to represent the minority of spending
using short-term claims and to have relatively low treated prevalence rates.
8
Annual beneficiary spending on nursing home care is available by type of stay, either
short-term or long-term. Spending is calculated from all possible sources, including Medicare,
Medicaid, private insurance, and out-of-pocket spending. As discussed, the majority of spending
for short-term stays is paid by Medicare, whereas spending on long-term stays is split mostly
between Medicaid and out-of-pocket spending. Spending was abstracted from the MCBS
personal summary file which provides total annual spending for short-term or long-term events
by patient. In the MCBS, spending for long-term patients represents both custodial and ancillary
costs (such as prescription drugs), though nursing home patients can still incur medical costs
outside of the nursing home. Short-term spending covers all medical care that a patient received
during the stay.
Spending by Disease
Past research on nursing home prices usually investigates per diem rates paid by different
payers. This paper investigates nursing home price growth in the 2000s by using a disease-based
price index, also known as a medical care expenditure (MCE) index. An MCE index defines the
price of nursing home care as annual patient spending for treatment of individual diseases, such
as total annual spending on the treatment of dementia. The best method for allocating spending
to exclusive diseases has not been determined and there is very little research available in the
area of nursing homes. The difference between methods is mostly driven by issues with
comorbidities (Dunn et al., 2015), which is especially significant in the nursing home population.
In a comprehensive analysis of U.S. spending by medical condition from 1996-2005, spending
on mental disorders and heart conditions were found to contribute the most to total spending
(Roehrig et al., 2009). While nursing home spending was part of the study, the public tables
show spending aggregated across all places of services (hospitals, hospice, etc.) and therefore
9
there are no estimates of nursing home spending by disease. This paper follows the methodology
in Roehrig et al. (2009) to allocate spending by disease for nursing home patients. This is done
by distributing annual patient spending equally to each disease indicated on a patient's survey or
SNF claim. Annual spending is then aggregated across all patients with each disease. In BEA’s
health satellite account, spending and patients are found using the first disease linked to each
medical event (called the “primary diagnosis” method). However, this method is not feasible to
allocate spending for long-term nursing home residents in the MCBS because disease
information is collected in an annual survey and is not linked to specific events. Although short-
term patients have individual claims diagnoses that can used in a primary diagnosis method,
most claims involve complex cases with multiple diagnoses listed. This differs from many of the
medical events from the BEA account, which often listed only one diagnosis, making the
primary diagnosis a reasonable method in those cases. Therefore, incorporating multiple diseases
appears to be necessary for accurate estimates of spending by disease for nursing homes patients.
A sensitivity analysis using regression techniques is also tested.
About half of the long-term survey conditions map to more than one CCS category. For
these conditions, spending is allocated using the distribution from SNF claims. For example, the
survey disease "hypertension" maps to CCS codes 98 and 99. An examination of the SNF claims
showed that between those two diagnoses, CCS 98 was indicated about 91% of the time.
Therefore, we allocate 91% of spending for a long-term resident with “hypertension” to CCS 98
and 9% of spending to CCS 99. Not all survey diseases are represented in short-term claims each
year, so in those cases the average share from all years was used for the allocation.
Events with zero spending and without diagnoses were considered incomplete and were
dropped from the analysis.
10
Medical Care Expenditure Indexes
An MCE index is calculated using annual per-patient spending on individual diseases as
the price (Aizcorbe, 2013). Estimates are aggregated to the annual level for each disease, and per
patient spending (i.e., price) is found by dividing total spending by number of patients. Price
growth measured with an MCE index reflects the change in patient costs for a fixed distribution
of diseases, regardless of payer. This differs from the conventional method of estimating nursing
homes price growth, which uses a Personal Consumption Expenditures (PCE) index. The PCE
index mainly relies on the Producer Price Index (PPI) for determining nursing home price
growth, available from the Bureau of Labor Statistics. In general, the PPI uses producer and
employer surveys to measure changes in per diem rates and for each individual nursing home
service by type of payer. Again, the advantage of measuring prices with a disease-based MCE
index is that the MCE picks up the effect of substitutions between types of insurance and
between different care settings. In the nursing home sector, when a home resident changes from
paying out-of-pocket to having Medicaid coverage, a common occurrence called “spending
down” (Wiener et al., 2013), an MCE index will captures this as a decline in price, whereas the
PPI and PCE index will not.
To obtain a larger sample size for the disease-based estimates and remain consistent with
the current presentation in the BEA health care satellite account (HCSA), spending and number
of patients are pooled together in two-year intervals. Therefore, the 2000 estimates include 1999
and 2000 data and the 2001 estimates include 2000 and 2001. The MCE subindex that we
calculate includes spending for all services at nursing homes. For each disease, d, the MCE index
for nursing home services, s, is found by dividing price in year t by the base year price:
11
,
=
,
,
where
,
= per patient expenditures by disease in year t (total expenditures by disease in year t
divided by number of patients with disease in year t) and
,
= per patient expenditures by
disease in year 1 (total expenditures by disease in year 1 divided by number of patients with
disease in year 1). The overall MCE index is found by summing the disease indexes, weighting
by share of base year expenditures:
=
(
,
,
)
where
,
= total expenditures for nursing home services for each disease in the base year
divided by overall base year expenditures for nursing home services.
Including the costs of all the care received in the nursing home in our index allows us to
properly account for any changes in the bundle of treatments used to treat medical conditions
(utilization) as a change in the cost of care (price). Previous research has shown that these
changes in utilization are potentially important in understanding the source of price growth for
many conditions (Dunn et al., forthcoming). Additionally, when residents spend down from
paying out-of-pocket to having Medicaid coverage, the MCE will capture this change as a price
drop.
Integrating the nursing home subindex with the BEA HCSA index requires we exclude
spending on services received by nursing home patients outside of the nursing home.
Numerically, the disease-level estimates of nursing home spending and patients are added to the
BEA estimates used to create health account index for all other medical care (physician and
hospital (inpatient, outpatient, emergency department) visits, prescription drugs, and home
12
health). This overall MCE index that combines other medical services (ns) with nursing home
care (s) is written:
= (
,
,
,
)
The combined index is essentially a weighted average of the nursing home subindex and
the BEA index. To the extent that nursing home patients do receive care in hospitals for the same
conditions that they report at the nursing home, we will inadvertently treat those two episodes as
if they were for different patients. This will double-count the number of patients suffering from
those conditions and will, thus, understate the average cost per patient. However, if this trend is
stable over time, it would not affect the price indexes. The likely impact to the overall result is
low given that long-term stays represent the majority of nursing home care and those residents
mostly stay in a nursing home for life. The extent of this problem is uncertain without linking
patients across the separate data series, an area outside of the scope of this paper.
RESULTS
The majority of nursing home spending for Medicare beneficiaries was attributable to
long-term care, accounting for 71% of total nursing home spending in 2009, down from 86% in
2000. The decline in the share of spending attributable to long-term care reflects a decrease in
the number of long-term nursing home residents during the period. This result corresponds with
the recent movement to keep patients at home and out of relatively expensive long-term care
facilities (Gleckman, 2009). Short-term nursing home care experienced a different trend as
utilization increased during the period. About 44% of beneficiaries with nursing home stays in
2009 were exclusively long-term residents (down from 54% in 2000), 40% were exclusively
13
short-term stays (up from 32%), and 16% of beneficiaries with a nursing home stay had both a
short-term and long-term stay (up from 14%). Annual average spending for long-term residents
was almost twice that for short-term patients (Figures 2.a & 2.b).
Long-term residents often received care for different diseases than short-term patients.
Almost half of spending for long-term residents was allocated to mental illnesses, which includes
conditions such as dementia, Alzheimer’s, and other mental conditions. Short-term patients had a
higher share of spending assigned to acute conditions, such as diseases of the musculoskeletal
system (e.g., back pain) and injury and poisoning (Table 1). For short-term residents, there were
slight differences in the distribution of disease spending when diagnoses from claims were used,
which contain up to nine diagnoses per event, versus when the survey was used, which contain
the first three claims diagnoses. Because claims provide more comprehensive information on
treated diseases than the survey, going forward results for short-term patients are reported using
claims.
The MCE indexes show different price growth for long-term and short-term stays during
the period. The average annual growth rate for the long-term MCE index was 1.4%, half the
growth rate for short-term stays (Figure 3). Growth is even lower (0.9%) when the two series are
combined into a single index. In this combined index, disease and spending information was
merged into a single episode for the approximately 15% of patients with both a short-term and
long-term stay. This resulted in relatively slower price growth for many conditions, picking up
the effect of shifting care from a short-term to long-term setting. When these patients are
excluded from the overall MCE index, price growth rises to 1.3%. In general, the slow growth in
the combined index is driven by long-term care spending, which represents most nursing home
spending. Price growth was slow or negative for many mental health conditions, which dominate
14
other disease categories in spending. Slow growth was apparent for many of the conditions with
large shares of spending, with the exception of diabetes (Table 2). For example, essential
hypertension, the disease with the most spending aside from mental conditions, averaged only
2.1% price growth annually in the combined index.
Adding nursing home spending to medical spending from BEA’s health care satellite
account (HCSA) reduces the average annual growth rates (AAGR) from 4.9% to 4.1% (Figure
4). The overall impact of incorporating nursing home spending into BEA’s new HCSA reflects
the slow growth in nursing home prices, driven by low and negative price growth seen in most
mental health conditions (Table 2). Although these diseases were the largest contributors to
driving down the growth, most conditions realized relatively slower price growth in nursing
homes compared to the HCSA. For example, average price growth for many cardiovascular
diseases, the most prevalent disease category in both the HCSA and among nursing home
patients (Figure 5), was much lower in nursing homes.
The slow growth in spending on mental conditions during the 2000s may indicate that
investments in specialized units for patients with cognitive conditions from the 1990s paid off,
by reducing growth in spending the following decade (Stewart et al., 2009). Alternatively, an
increased focus on nursing home quality of care through various pay-for-performance programs
may have diverted spending to aspects of care measured in the programs, most of which were not
directly related to treating mental conditions (Werner & Asch, 2005). Unfortunately, the MCBS
survey does not provide comprehensive details on the types of services provided to long-term
nursing patients. Additional research is needed to understand the specific factors leading to the
flat, and sometimes negative, price growth for mental conditions.
15
Nursing home price growth using the official PCE index was 4 times higher than the
1.0% growth rate found using disease-based price indexes (re-basing to 2009 to match the PCE
index slighted increased the growth rate). The growth rate for the nursing home PCE index is
identical to growth in the producer price index (PPI) for nursing home care. The difference in
growth rates between the MCE and PCE indexes reflects the different definitions of output.
Specifically, the MCE index tracks growth in prices for the treatment of diseases, whereas the
PCE index (and PPI) tracks prices for nursing home services by the type of payer. The PCE
index relies on producer and employer surveys to determine price growth in nursing homes by
type of payer, whereas an MCE index tracks the price of spending on the treatment of diseases,
regardless of who pays. Aside from not capturing the Medicaid “spend down” phenomenon, the
PCE index also does not capture cuts in service or shifts happening within the Medicaid program
over time. In the early 2000s, many states began working with private insurers to lower spending
by enrolling Medicaid recipients into managed care programs (Galewitz, 2011). If the PCE index
is not picking up this movement into managed care programs by long-term Medicaid recipients it
will not capture the resulting drop in price for nursing home care.
Sensitivity Analyses
The first sensitivity analysis determines if the MCBS nursing home population is a
reasonable proxy for the entire universe of nursing home patients by comparing results from the
MCBS with another source for nursing home data, the National Nursing Home Survey (NNHS).
Data are available from the NNHS for four years: 1995, 1997, 1999, and 2004. The advantage of
the NNHS is that it covered all nursing home residents, not just Medicare beneficiaries, as with
the MCBS. However, the survey was changed in the late 2000s to one that covers only facility
characteristics and no longer contains disaggregated resident-level information necessary to
16
calculate spending by disease. For the comparison, the MCBS data were transformed to per diem
rates to correspond with the format of the NNHS file, which does not provide annual totals for
patients. Per diem prices were relatively similar for NNHS and MCBS long-term residents for
the four available years, though they diverge slightly in 2004 (Table 3). Short-term MCBS rates
were much higher than the NNHS and long-term MCBS rates, but this is expected because
patient severity and resources used are generally higher for care after a hospital stays. The NNHS
documentation acknowledges long-term residents are over-represented due to its survey design
(Centers for Disease Control and Prevention, 2009), and these results verify this bias. Spending
by disease was also calculated using the NNHS data, following similar methods as with the
MCBS data. The NNHS data had a similar distribution as the MCBS for most ICD-9 chapters
(not shown). An important exception is that the NNHS includes all ICD-9 chapters, whereas the
MCBS survey for long-term patients does not ask about less prevalent diseases and therefore is
not a comprehensive distribution of diseases. As a result, the NNHS allocates about 15% of total
spending to conditions not covered in the MCBS survey of long-term residents.
The second sensitivity analysis tests if the spending by disease estimates are sensitive to
the method used by using another method to allocate spending. Following previous research on
dividing spending to diseases in the absence of complete claims diagnoses, we use regression
coefficients to allocate spending to individual conditions (Hall & Highfill, 2013). To do this,
patient spending is first regressed on the conditions a patient is diagnosed with, then spending is
allocated to each diagnosed conditions in proportion to the regression coefficient (Trogdon et al.,
2008). As with the primary method, total annual spending is found by summing across all
patients for each disease. The results showed a similar pattern of spending for most diseases.
However, the regression allocated a significantly larger share of spending to mental conditions.
17
Using this method of spending, the AAGR would be even lower than the main result. Given the
relatively small sample sizes for some of the diseases, regression coefficients often fluctuate
from year to year, causing issues with reliability.
Despite the care taken in designing this study, there are important caveats to consider
when interpreting the results. Primarily, the data used are subject to a few limitations. The
MCBS only covers those with Medicare and does not represent the entire population of nursing
home patients. While the NNHS data show that 11% of nursing home patients in 2004 were
under the age of 65, this may not be a good metric of Medicare coverage because those under
age 65 can be covered if they have an eligible disability. About 7% of the MCBS nursing home
patients in 2009 were under age 65, and the majority of those were covered by Medicaid. Since
the vast majority of nursing home patients are also covered by Medicare, the MCBS appears to
be a reasonable proxy for the nursing home populations. The MCBS data do not contain
comprehensive details on the services provided to nursing home residents, making an
examination of intensity of service difficult. Therefore, it is unclear whether the slow growth in
nursing home prices reflects a decrease in the number of services for residents or if something
else is occurring. Additionally, because we assume that nursing home patients do not receive
non-nursing home care in our subindex, we cannot properly account for any treatment shifts
between nursing homes and other settings (e.g., home health care). To the extent that these shifts
are occurring to lower-cost settings, our index will overstate true costs.
While both methods for allocating spending ended up with similar overall results, it is not
clear which technique is most accurate and both have strengths and weaknesses. For example,
the method of dividing spending equally across patients’ diagnoses allocates a large portion of
spending to hypertension because that is a highly-prevalent disease among the elderly. If a
18
person has both Alzheimer’s and hypertension, spending is allocated equally between the two
diseases, even though it seems more likely that treating the Alzheimer’s would be a much higher
expense. On the other hand, the regression method relies on small sample sizes for many
diseases and produces variable coefficient estimates from year to year. There is also some
disagreement over whether it is appropriate to divide spending on nursing home care into
exclusive disease categories, since so many patients are admitted due to general frailty that may
not be attributable to a specific disease or condition (National Research Council, 2010). Whether
or not nursing home spending will be treated separately from other medical spending in the
HCSA is still not settled. Nonetheless, the contribution of this paper is to provide a first estimate
of what BEA’s new disease-based health care satellite account may look like when nursing home
spending is included.
CONCLUSION
Including nursing home spending into BEA’s health care satellite account (HCSA)
provides a more comprehensive picture of the U.S. health care sector. Spending for nursing
home care represents almost 9% of PCE for health care services (U.S. Bureau of Economic
Analysis, 2015), but is not currently included in the HCSA. When nursing home spending is
added to BEA’s medical care spending estimates in the HCSA, the AAGR for the new combined
MCE price index declines almost a percentage point. Additionally, the rate of overall health care
spending is estimated to grow slower than originally thought when nursing home spending is
included. Mental conditions are responsible for the largest share of nursing home expenditures
and have much slower growth rates than in the HCSA, however most diseases showed relatively
slower price growth for nursing homes. The weighted MCE index of nursing home and other
medical spending provides a general idea of the impact of incorporating nursing home spending
19
into the HCSA, but a more rigorous method is necessary to deal with potential double-counting
of patients in both surveys. An “optimal” account would follow individual patients throughout
their lives to track their health service utilization over time, including conditions treated,
insurance status, place of service, and spending. No single data source contains this information
for a nationally-representative sample of people, though techniques using survey weights have
some potential to circumvent issues with representativeness (Dunn et al., 2015).
Adding nursing home expenditures to the HCSA will provide a more comprehensive
picture of health care spending in the U.S. However, significant methodological and data
challenges must first be addressed before nursing homes can be fully incorporated into BEA’s
new health care satellite account.
20
REFERENCES
Aizcorbe, A. (2013). Recent research on disease-based price indexes: Where do we stand?
Survey of Current Business 93 (July), 9–13.
Aizcorbe, A., & N. Nestoriak. (2011). Changing mix of medical care services: Stylized facts and
implications for price Indexes. Journal of Health Economics 30, no. 3 (May): 568–574.
Berndt, E., D. Cutler, R. Frank, Z. Griliches, J. Newhouse, & J. Triplett. (2000). Medical care
prices and output. Handbook of Health Economics, edited by Anthony J. Culyer, and
Joseph P Newhouse, 119–180. Amsterdam, The Netherlands: North Holland.
Centers for Disease Control and Prevention. (2009). The National Nursing Home Survey: 2004
Overview. Vital and Health Statistics Series 13, Number 167 (June 2009). Washington
D.C.
Centers for Medicare & Medicaid Services. (2013). Nursing home data compendium: 2012
edition. Retrieved April 29, 2015 from http://www.cms.gov/Medicare/Provider-
Enrollment-and-Certification/CertificationandComplianc/NHs.html.
Centers for Medicare & Medicaid Services. (2015). Medicare 2015 costs at a glance. Retrieved
April 29, 2015 from http://www.medicare.gov/your-medicare-costs/costs-at-a-
glance/costs-at-glance.
Clement, JP, Bazzoli, GJ, & Zhao, M. (2012). Nursing home price and quality responses to
publicly reported quality information. Health Services Research 47: 86-105.
Dunn, A, EB Liebman, A Shapiro (Forthcoming) Implications of utilization shifts on medical-
care price measurement. Health Economics.
21
Dunn, A., Rittmueller, L., & Whitmire, B. (2015). Introducing the new BEA health care account.
Survey of Current Business 95 (January), 1-21.
Galewitz, P. (2011). States turn to private insurance companies for managed care. USA Today.
Last updated Feb. 21, 2011. Retrieved April 14, 2015 from
http://usatoday30.usatoday.com/money/industries/health/2011-02-21-
longtermcare21_ST_N.htm.
Gleckman, H. (2009). The death of nursing homes. Kaiser Health News. Retrieved April 29,
2015 from
http://www.kaiserhealthnews.org/Columns/2009/September/092809Gleckman.aspx
Grabowski, D., Stevenson, D., & Cornell, P. (2012). Assisted living expansion and the market
for nursing home care. Health Services Research, 47(6), 2296-2315.
Hall, A. & Highfill, T. (2013). A regression-based medical care expenditure index for Medicare
beneficiaries. U.S. Bureau of Economic Analysis Working Paper: WP2013-04.
Washington, D.C.
Medicare Current Beneficiary Survey. (2012). Calendar year 2009 cost and use file, technical
appendix 1. U.S. Department of Health and Human Services.
Mukamel, D. B., Spector, W. D., Zinn, J., Weimer, D. L., & Ahn, R. (2010). Changes in clinical
and hotel expenditures following publication of the nursing home compare report card.
Medical Care, 48(10), 869-874.
National Center for Health Statistics. (2014). Health, United States, 2013: With special feature
on prescription drugs. Hyattsville, Maryland.
22
National Research Council. (2010). Accounting for health and health care: Approaches to
measuring the sources and costs of their improvement. Washington, DC: The National
Academies Press.
Northwestern Mutual. (2013). Northwestern Long Term Care Insurance Company. Retrieved
April 29, 2015 from http://www.northwesternmutual.com/about-northwestern-
mutual/our-company/northwestern-mutual-subsidiaries/northwestern-long-term-care-
insurance-company.aspx.
Paradise, Julia. (2015). Medicaid moving forward. Issue Brief. The Henry J. Kaiser Family
Foundation. March 2015.
Roehrig, C., Miller, G., Lake, C., & Bryant, J. (2009). National health spending by medical
condition, 1996–2005. Health Affairs, 28(2), w358-w367.
Rosenberg, M. A., & Browne, M. J. (2001). The impact of the inpatient prospective payment
system and diagnosis-related groups: a survey of the literature. North American Actuarial
Journal, 5(4), 84-94.
Stewart, K. A., Grabowski, D. C., & Lakdawalla, D. N. (2009). Annual expenditures for nursing
home care: Private and public payer price growth, 1977–2004. Medical Care, 47(3), 295.
Trogdon, J., Finkelstein, E., and Hoerger, T. (2008). Use of econometric models to estimate
expenditure shares. Health Services Research 43, no. 4 (August): 1,442–1,452.
U.S. Bureau of Economic Analysis (2015). National Income and Product Table 2.3.5. Personal
Consumption Expenditures by Major Type of Product. Updated June 24, 2015. Retrieved
June 29, 2015 at http://www.bea.gov/iTable/index_nipa.cfm.
23
U.S. Department of Health and Human Services. (2010). Multiple chronic illnesses-A strategic
framework: Optimum health and quality of life with multiple chronic conditions.
Washington, D.C.
Werner, R. M., & Asch, D. A. (2005). The unintended consequences of publicly reporting
quality information. JAMA, 293(10), 1239-1244.
Wiener, J. M., Anderson, W. L., Khatutsky, G., Kaganova, Y., & O’Keeffe, J. (2013). Medicaid
spend down: Implications for long-term services and supports and aging policy. Long
Beach, CA: The Scan Foundation. Retrieved May 6, 2015 at http://www.
thescanfoundation. org/sites/thescanfoundation. org/files/tsf_ltc-financing_medicaid-
spend-down-implications_wiener-tumlinson_3-20-13_0. pdf.
Zuckerman, S., Williams, A. F., & Stockley, K. E. (2009). Trends in Medicaid physician fees,
2003–2008. Health Affairs, 28(3), w510-w519.
24
TABLES
Table 1. Spending by Disease for Nursing Home Patients with Medicare by Type of Stay, 2004
(%)
Disease Chapters Long-term
Short-term,
Survey
Diagnoses
Short-term,
Claims
Diagnoses
Long-term +
Short-term,
Claims
Diagnoses
Infectious and parasitic diseases - 0.7 0.9 0.2
Neoplasms 1.4 2.2 1.9 1.5
Endocrine, nutritional, metabolic
diseases & immunity diseases
6.7 7.4 10.0 7.4
Diseases of the blood and blood-
forming organs
- 2.0 3.7 0.8
Mental illness 42.1 4.0 7.0 34.8
Diseases of the nervous system and
sense organs
4.4 7.8 6.5 4.9
Diseases of the circulatory system 31.1 21.4 24.5 29.8
Diseases of the respiratory system 3.2 8.5 6.5 3.9
Diseases of the digestive system - 4.1 6.1 1.3
Diseases of the genitourinary system 0.4 5.0 5.1 1.4
Diseases of the skin and subcutaneous
tissue
- 2.4 2.4 0.5
Diseases of the musculoskeletal
system & connective tissue
9.5 9.3 12.0 10.0
Injury and poisoning 1.0 14.6 8.4 2.5
Other conditions - 9.4 2.0 0.4
Residual, unclassified, E Codes - 1.2 2.7 0.6
Note: Pregnancy-related conditions not shown (0% of spending).
25
Table 2. Medical Care Expenditure Indexes: Decomposition between 2000 & 2009 (%)
Condition
2000 Nursing
Home
Expenditure
Share
Average Annual Growth Rate (AAGR)
Nursing Homes
Health Care Satellite
Account*
Price
Prevalence
Price
Prevalence
Senility and organic mental
disorder
14.2 1.1 -0.0 10.0 4.0
Other mental conditions 13.5 -2.0 2.9 1.3 0.4
Essential hypertension 9.9 2.1 2.7 3.0 5.3
Osteoarthritis 4.7 1.4 0.4 5.2 4.9
Acute cerebrovascular
disease
4.5 2.7 -3.4 3.5 0.8
Congestive heart failure,
non-hypertensive
4.3 1.3 0.3 4.6 0.7
Anxiety, somatoform, and
dissociative disorders
3.6 -0.1 5.8 5.2 7.1
Diabetes mellitus without
complication
3.6 3.0 5.8 1.0 7.4
Coronary atherosclerosis
and other heart disease
3.5 1.4 0.3 2.2 1.9
Osteoporosis 3.1 1.9 3.1 5.7 2.3
Note: Conditions where 2000 nursing home expenditure share > 3%.
*Spending on hospitals (inpatient, outpatient, and emergency department), physician services,
prescription medications, and home health services.
26
Table 3. Average Daily Spending for Nursing Home Patients: Data Source Sensitivity Analysis
($)
Year
Medicare Current
Beneficiary Survey,
Short-term Residents
Medicare Current
Beneficiary Survey,
Long-term Residents
National Nursing Home
Survey
1995 481 105 105
1997 598 111 119
1999 541 112 129
2004 753 130 169
27
FIGURES
Other
11%
Private
10%
Figure 1.a Short-term Nursing Home Spending
by Payer, 2004
Medicare
76%
Medicaid 4%
Private
40%
Other
9%
Figure 1.b Long-term Nursing Home Spending
by Payer, 2004
Medicaid
51%
$21 billion
$78 billion
28
Long-term
Short-term
Both stays
400
600
800
1,000
1,200
1,400
1,600
1,800
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
(000s)
Figure 2a. Number of Nursing Home Patients by Type of Stay
Long-term
Both stays
Short-term
0
10,000
20,000
30,000
40,000
50,000
60,000
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
$
Figure 2b. Nursing Home Annual Per-patient Spending by Type of Stay
29
Short-term (2.8%)
Long-term (1.4%)
Combined (0.9%)
0.95
1.05
1.15
1.25
1.35
1.45
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Figure 3. Nursing Home Medical Care Expenditure Indexes (Average Annual
Growth Rate)
30
Nursing Homes
(0.9%)
Health Care
Satellite Account
(HCSA)
(4.9%)
HCSA + Nursing
Homes
(4.1%)
0.95
1.05
1.15
1.25
1.35
1.45
1.55
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Figure 4. Medical Care Expenditure Indexes (Average Annual Growth Rate)
31
Endocrine,
nutritional, and
metabolic diseases
and immunity
disorders
Mental illness
Diseases of the
nervous system
and sense organs
Diseases of the
circulatory system
Diseases of the
respiratory system
Diseases of the
musculoskeletal
system and
connective tissue
0%
5%
10%
15%
20%
25%
30%
35%
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Figure 5. Prevalence for Nursing Home Patients (Average Prevalence > 5%)
