Cite this article as:

At the Intersection of Health, Health Care and Policy
Cite this article as:
Sheila Smith, Joseph P. Newhouse and Mark S. Freeland
Income, Insurance, And Technology: Why Does Health Spending Outpace Economic
Growth?
Health Affairs, 28, no.5 (2009):1276-1284
doi: 10.1377/hlthaff.28.5.1276
The online version of this article, along with updated information and services, is
available at:
http://content.healthaffairs.org/content/28/5/1276.full.html
For Reprints, Links & Permissions:
http://healthaffairs.org/1340_reprints.php
E-mail Alerts : http://content.healthaffairs.org/subscriptions/etoc.dtl
To Subscribe: http://content.healthaffairs.org/subscriptions/online.shtml
Health Affairs is published monthly by Project HOPE at 7500 Old Georgetown Road, Suite 600,
Bethesda, MD 20814-6133. Copyright © 2009 by Project HOPE - The People-to-People Health
Foundation. As provided by United States copyright law (Title 17, U.S. Code), no part of Health Affairs
may be reproduced, displayed, or transmitted in any form or by any means, electronic or mechanical,
including photocopying or by information storage or retrieval systems, without prior written permission
from the Publisher. All rights reserved.
Not for commercial use or unauthorized distribution
Downloaded from content.healthaffairs.org by Health Affairs on October 1, 2014
by guest
O v e rv i e w
Income, Insurance, And
Technology: Why Does Health
Spending Outpace Economic
Growth?
Insurance’s effect may be less than in the past, particularly in its
impact on the use and spread of new medical technologies.
by Sheila Smith, Joseph P. Newhouse, and Mark S. Freeland
ABSTRACT: A broad consensus holds that increased medical capability—technology—is
the primary driver of health spending growth. However, technology does not expand independently of historical context; it is fueled by rising incomes and more generous insurance
coverage. We estimate that medical technology explains 27–48 percent of health spending
growth since 1960—a smaller percentage than earlier estimates. Income (gross domestic
product, or GDP) growth plays a critical role, primarily through the actions of governments
and employers on behalf of pools of consumers. The contribution of insurance is likely to
differ, with less of a push from increasing generosity of coverage and more of a push from
changes in provider payment. [Health Aff (Millwood). 2009;28(5):1276–84; 10.1377/
hlthaff.28.5.1276]
A
f t e r t e m p o r a r i ly s l o w i n g i n t h e m i d - 19 9 0 s , the percentage of
U.S. gross domestic product (GDP) devoted to health care has resumed its
apparently inexorable rise.1 The troubling implications for the long-term
viability of the current U.S. system of financing and providing health services are
all too clear. The ability to formulate appropriate policies to reduce health spending growth requires understanding the major factors driving that growth—an urgent task, given the difficulty of making sharp reductions in a short time.
Changing medical technology is one of the few factors that can potentially explain persistently high growth in medical spending over time and across many
countries.2 Indeed, the dominant role of technology as a driver of spending has become a truism in health economics.3 The conclusion that technological change explains much of the growth rests on a macroeconomic approach that seeks to esti-
Sheila Smith ([email protected]) and Mark Freeland are economists in the Office of the Actuary, Centers for
Medicare and Medicaid Services, in Baltimore, Maryland. Joseph Newhouse is the John D. MacArthur Professor of
Health Policy and Management at Harvard University in Boston, Massachusetts.
1276
September/October 2009
DOI 10.1377/hlthaff.28.5.1276 ©2009 Project HOPE–The People-to-People Health Foundation, Inc.
Downloaded from content.healthaffairs.org by Health Affairs on October 1, 2014
by guest
Sp e ndi ng
&
G D P
mate the contribution of known factors to health spending growth and assumes
that most of the unexplained residual growth is attributable to technology.2
Joseph Newhouse’s 1992 and 1993 papers4, 5 are well-known examples of this approach. He concluded that if technology had been constant, demographic changes,
income growth, and insurance growth would have accounted for “well under
half—perhaps under a quarter”—of the increase in medical care spending between 1940 and 1990. An expanding number of empirical studies on specific medical technologies support the key role assigned to technology but are largely agnostic on the magnitude of its effect in the aggregate.6
In this paper we update Newhouse’s 1992 estimate with seventeen more-recent
years of data and revise his methodology to reflect subsequent research. The contribution of technology by itself to health spending growth is smaller than his earlier estimate suggests. But the interrelationships among technology, income, and
insurance are strong; the burgeoning of medical technology since World War II
did not occur in isolation but, rather, in the context of growing economies and
health care financing institutions that facilitated technology’s development and
diffusion. These interrelationships make it difficult to assign specific quantitative
magnitudes to each factor, although we seek to provide ranges.
In our work, growth in aggregate income (GDP) stands out as an important
driver of health care cost growth, despite the small role income plays in insured
households’ health care decisions. Theory and recent empirical work suggest that
the generosity of insurance coverage has also stimulated technological change.
Looking forward, its role could well differ from its role in the past: there is less
room for rising generosity of coverage and the probability of a continued evolution
in the structure and constraints.
Study Data And Methods
The primary objective of the original Newhouse estimate was estimating welfare loss—the costs to society from excessive spending on medical care resulting
from insurance—and to argue that the economics literature at that time overestimated that loss. In this work we changed the objective from estimating welfare
loss to estimating the relative contribution of various drivers of long-term growth
in aggregate health spending. This prompted two changes in methodology.
First, we changed the method used to estimate the importance of each causal
factor as a driver of spending growth. This new method7 shows that several factors
contributing to growth, such as economic prosperity, rising medical prices, and
expanding insurance coverage, played major roles; it implies a smaller role for
medical technology than in the original Newhouse estimate.
Second, to approximate the magnitude of the aggregate income effect, we relied
on estimates based on how health spending varies across countries and time. The
consensus that technology is the major driving factor was built using estimated
income and insurance effects at the household level, especially estimates from the
H E A L T H A F F A I R S ~ Vo l u m e 2 8 , N u m b e r 5
Downloaded from content.healthaffairs.org by Health Affairs on October 1, 2014
by guest
1277
O v e rv i e w
RAND Health Insurance Experiment (HIE). These implied that variation in income, insurance, and demographics across households explained little of the
growth in aggregate health spending over time.8, 9 The most plausible inference
was that changing technology accounted for much of the growth.
Since that time, there has been growing recognition that the responsiveness of
health spending to income and insurance is larger at the macro or country level
than at the household level.10–16 The large residual implied by using householdlevel income estimates overstates the pure contribution of technology to aggregate
health spending growth. Insurance—as well as income—likely has effects that are
larger at the macro level than household variation suggests. However, we continue
to rely on the HIE for this purpose, because the uncertainty on this question precludes generalization of the size of this effect at the macro level.
The gap between the micro and the macro estimates of the effects of income is
partly attributable to measurement problems and aggregation issues, but it likely
reflects economic factors as well. In particular, the rate of technological innovation is influenced by the size of the market, which in turn is influenced by income
and insurance.14, 17 Using macro estimates of the magnitude-of-income effects is
more appropriate for analysis of spending growth at the national level.10 However,
this does make it more difficult to control accurately for the effect of all factors
that may influence growth and to estimate the interactions among them.
The effects of income, insurance coverage, and medical technology on health
spending clearly form a critical nexus. In the interest of brevity, we focus our discussion on these three factors and relegate much of the supporting technical work
to an appendix, including our estimates of demographic effects.7 We do not quantify the role of administrative costs, the health status of the population, supplierinduced demand, and potential inefficiency due to professional liability concerns.
Spending attributable to most of these factors is simply too small to account for
much of the historical growth in health spending.2 Moreover, these factors are not
plausible as consistent drivers of spending growth over all countries in the Organization for Economic Cooperation and Development (OECD) for five decades. To
the extent that these factors have caused spending to grow, however, the impact
will be reflected in our estimated effects and residual, which will therefore overstate the contribution of income, insurance, and technological change.
n Medical price inflation. Although it seems natural to use medical price indexes to estimate the direct effect of medical price inflation on spending, we have
not done so. The available historical price indexes are widely believed to overstate
growth in medical prices by a substantial margin. Rather, we assumed that prices
track the average cost of production, determined by the price of inputs and by production efficiency (total factor productivity), and we used estimates of those two
terms to derive a measure of medical (output) price growth.18 We estimated input
price inflation for medical care relative to the entire economy using quality-adjusted
measures of volume and price of capital, labor, energy, and materials.19 Our upper-
1278
September/October 2009
Downloaded from content.healthaffairs.org by Health Affairs on October 1, 2014
by guest
Sp e ndi ng
&
G D P
bound assumption for total factor productivity was 0.8 percent (the economywide
average), and our lower-bound assumption was zero.
This methodology implies that rising medical prices contributed 0.2–0.9 percentage points to annual growth in real per capita health spending during 1960–
2007, or 5–19 percent of the total (Exhibit 1). The range reflects the alternative assumptions on productivity in health care.
n Insurance coverage. Quantifying the effects of insurance on aggregate spending is difficult. Insurance coverage influences health spending through three channels: the proportion of the population with insurance; the depth of coverage in insurance contracts; and the level and structure of reimbursement from insurers, both
public and private, to providers. We accounted for all of these channels based on average out-of-pocket share of personal health spending paid by the consumer, which
declined from 55 percent in 1960 to 14 percent in 2007. If one uses an estimated insurance elasticity of –0.2 from the HIE, the net expansion in insurance coverage accounted for 10.8 percent of the growth in real per capita health spending.9
This relatively minor role for insurance coverage as a driver of growth based on
a household-level elasticity estimate may well be low, however. Because insured
patients bear only a fraction of the costs of new technologies and because U.S. reimbursement was largely passive over much of the period, incentives for inefficiently broad application of new technologies and thus higher cost growth rates
(dynamic moral hazard) were present.20, 21 Dynamic moral hazard implies that
even an unchanging rate of insurance coverage can drive chronic growth in medical spending because of persistent overuse. However, this incentive may be reduced through methods of payment that cut the link between payment and volume of care (for example, prospective payment or other methods of bundling).
Empirically, Amy Finkelstein concludes that the expansion of insurance coverage
resulting from the introduction of Medicare and Medicaid eventually had sizable
market-level effects on health spending over the years following its introduction.11
Changes in coverage affected a large share of the population and so affected capital
investment (for a given technology), standards of care, and the nature of technologies that were subsequently introduced and their rate of diffusion.11, 17, 22
Nonetheless, the magnitude of the effects of market-level changes in insurance
coverage (as distinct from household-level changes) on health spending remains
highly uncertain. For that reason, we used the 10.8 percent value based on household responses to insurance as the portion of cost growth due to insurance expansion; this means that some of the cost growth that we attributed to income and
technology may actually be attributable to insurance.
n Income effects. Income (real per capita GDP) is a critical factor in determining how much nations spend on medical care; it consistently explains around 90
percent of variation in real health spending across countries and time. Recent estimates tend to find a macro-level income elasticity of about 1.0, implying that health
spending moves in tandem with GDP.13, 23 However, the raw or unadjusted elasticity
H E A L T H A F F A I R S ~ Vo l u m e 2 8 , N u m b e r 5
Downloaded from content.healthaffairs.org by Health Affairs on October 1, 2014
by guest
1279
O v e rv i e w
EXHIBIT 1
Percentage Of Average Annual Growth In Real Per Capita Health Spending
Attributable To Various Causal Factors, 1960–2007
Income elasticity
Medical care productivity =
economy average
Medical care productivity = zero
(1)a
(3)a
(2)b
(4)b
0.6
0.9
0.6
0.9
Income effects
Relative medical price inflation
Demographic effects
Change in insurance coverage
28.7%
5.0
7.2
10.8
43.1%
5.0
7.2
10.8
28.7%
18.8
7.2
10.8
43.1%
11.5*
7.2
10.8
Technology
Technology-income interaction
Technology residual
48.3
27.4
26.4
33.9
27.4
9.9
34.6
27.4
12.8
27.4*
27.4*
0.0*
100.0
100.0
100.0
Total
100.0
SOURCE: Authors’ calculations; see below.
NOTES: Minimum contribution of technological change is based on the difference between the unadjusted expenditure
elasticity and the expenditure elasticity with controls for common trend in health spending across Organization for Economic
Cooperation and Development (OECD) countries (fixed effects) over time. The contribution of relative medical price inflation
has been adjusted downward for this scenario, to prevent the contribution of technology from falling below this minimum.
Income is real per capita disposable personal income, five-year moving average. Relative medical price inflation is the Centers
for Medicare and Medicaid Services (CMS) personal health care price deflator, adjusted by constant trend to adjust for bias
consistent with assumptions. Insurance coverage is the National Health Expenditures Accounts (NHEA) out-of-pocket spending
as share of total personal health care spending. For all columns, estimates for the contribution of insurance and demographics
remain the same; only income and price effects vary: demographic estimate based on CMS age-sex indexes, which are largely
based on data from the Medical Expenditure Panel Survey (MEPS) with supplemental data for institutional settings. Insurance
coverage is based on arc elasticity from the RAND Health Insurance Experiment. TFP is total factor productivity. Factor price
inflation from Ho/Samuels database.
a
Expenditure elasticity = 1.6; income elasticity = 0.6; price elasticity = –0.2.
b
Expenditure elasticity = 1.6; income elasticity = 0.9; price elasticity = –0.2.
between real per capita health spending and real per capita GDP is much higher—
about 1.4–1.7.24 This unadjusted value, which we call an expenditure elasticity, reflects
not only a pure income effect but also other factors affecting health spending that
are correlated with real per capita GDP. This arguably includes a good part of the
impact of technology, medical prices, and insurance.
We used the raw expenditure elasticity to derive an estimate of a pure income
effect, using a standard model to estimate an adjusted elasticity. That model is
simple: real per capita health spending is estimated as a function of real per capita
GDP and a set of variables that capture variation in demographic variables and in
specific characteristics of the various countries’ health systems. In addition, the
model includes two-way fixed effects (meaning a set of yearly dummy variables
that measure trends common to all countries over time, and also a constant for
each country that is unvarying over time).
This model is appropriate for estimating income elasticity but is less helpful for
projecting health spending growth. For forecasting purposes, using this model
simply shifts the forecasting problem to one of forecasting future period (year) ef-
1280
September/October 2009
Downloaded from content.healthaffairs.org by Health Affairs on October 1, 2014
by guest
Sp e ndi ng
&
G D P
fects. Nonetheless, with an additional assumption, this model is useful for our
purposes. That additional assumption concerns the interpretation of the yearly
dummy variables that measure changes in spending that are common across countries. We assumed that these are most plausibly attributed to the common effect of
medical technology. They also include the effects of any common income and insurance trends, but because a country’s deviation from the common income trend
is included in the model and because common insurance trends are probably
small, we believe that the bulk of this measure is the common effect of technology,
and probably accounts for most of the effect. Micro-level evidence on the diffusion
of specific technologies supports this view, because early adoption and rapid diffusion of new technologies are frequently linked to real per capita GDP.25 Interpreting these period effects as technology and taking them out of the expenditure
elasticity leaves a remaining (partial) expenditure elasticity of 1.0 for 1960–2007.
To derive the pure effect of income, we also needed to account for the role
played by price. William Baumol’s well-known “cost disease” model suggests that
productivity growth may be intrinsically low (and price inflation relatively high)
for medical services, which are necessarily highly customized.26 To the extent that
this model holds, it implies that higher-income countries will have higher relative
medical prices, which could account for part of the relationship captured in the
expenditure elasticity. We approximated the potential magnitude of this interaction effect based on the same upper and lower bounds for total factor productivity
discussed earlier. We found that this effect lowers the expenditure elasticity by a
further 0.2–0.4. The remaining 0.6–0.9 of the expenditure elasticity we interpret
as the income effect.27 This estimate implies that 27–43 percent of growth in real
per capita health spending is attributable to income growth (Exhibit 1).
We explored whether we needed to make a further adjustment for the role of insurance using the out-of-pocket spending share as a proxy for coverage variations
across countries. Theory is ambiguous on the question of whether higher incomes
will raise or lower preferred levels of insurance coverage, and we found an (insignificant) effect that did not materially influence the estimated income elasticity.27
Discussion
Real per capita health spending grew roughly by a factor of 9 during 1960–
2007—an annual growth rate of 4.8 percent. Spending on new medical technologies accounted for 27–48 percent of this amount. Income accounted for a roughly
similar amount—another 29–43 percent. Economic theory suggests that these
two forces reinforced each other (Exhibit 1).
Increases in unit price accounted for 5–19 percent of the increase in health
spending, depending on how productivity changed in medical care. The smaller
the productivity increase in medical care relative to the economy as a whole, the
greater this value. Demographics appear to have played a small role in the historical growth of spending7 but will loom larger with the aging of the baby boomers.
H E A L T H A F F A I R S ~ Vo l u m e 2 8 , N u m b e r 5
Downloaded from content.healthaffairs.org by Health Affairs on October 1, 2014
by guest
1281
O v e rv i e w
Why is the aggregate relationship so strong between income and spending
when the income effect at the household level is small? A main reason is insurance;
a household only pays for care net of insurance, but our results imply that countries have mechanisms to adapt their spending to their incomes.
n Applicability of the model to U.S. experience. It is natural to ask how well
our model, using data from twenty-three countries, tracks U.S. historical experience. To do so we followed the standard Centers for Medicare and Medicaid Services (CMS) assumption that the macro-level response to a change in demand for
medical care occurs with moving average lag of five years; that is, we calculated the
average of the prior five annual values for income and used that average to predict
current health spending for a given year.28 Changes to the structure of insurance
contracts between (public and private) insurers and providers and the regulatory
infrastructure require a lengthy decision-making process, followed by a period of
implementation. With this assumption, our model tracks growth in real per capita
health spending remarkably well over time (Exhibit 2). That this estimate, rooted in
empirical estimates of causal factors that explain international variation in spending, can so effectively fit historical growth in U.S. health spending adds plausibility
to our estimates of the roles of each of the major drivers of cost growth.
EXHIBIT 2
Contribution Of Medical Technology And Non-Technology Factors To Growth In Real
Per Capita Health Spending, 1964–2007
Percent change
9
Real per capita health spending predicted by decomposition
8
7
Real per capita health spending
6
5
4
3
2
Contribution of medical
technology (smoothed)
1
Contribution of non-technology
factors
0
1965
1970
1975
1980
1985
1990
1995
2000
2005
SOURCE: Authors’ calculations of data from the Centers for Medicare and Medicaid Services.
1282
September/October 2009
Downloaded from content.healthaffairs.org by Health Affairs on October 1, 2014
by guest
Sp e ndi ng
&
G D P
n
Speculation about the future. Our model suggests that the unusual severity
of the current recession will reduce spending growth in the near term by an amount
roughly comparable in magnitude to that of the managed care era of the 1990s. Unfortunately, the model says nothing about how that reduction may be brought
about.
The rising generosity of insurance accounted for a relatively small share of
growth based on household estimates, but its role as a driver of technology was
important. It is likely to be less so in the future. Of course, the potential exists for a
substantial near-term stimulus to the demand for medical care if coverage is expanded to the currently uninsured population. Over the longer term, however, insurance is not likely to exert as large a positive effect on spending as it has historically, because it already covers 86 percent of personal health care spending, up
from 45 percent in 1960.29 Moreover, insurance reimbursement in both public and
private insurance has evolved from the passive indemnity fee-for-service coverage
of 1960 to much more active roles for both private and public insurers. Networks
of providers are now almost universal in private insurance. Diagnosis-related
groups (DRGs) for prospective payment were introduced by Medicare and then
adopted by many private insurers, and there is now discussion of further broadening the basis of payment to episodes of care. In short, insurance may play a less
stimulative role than in the past because reimbursement could be less passive and
because the out-of-pocket spending share will not fall by the amount it has in the
past.
Income growth will continue to drive a rising health share of GDP in decades to
come, as spending on new medical technologies continues to increase more rapidly than incomes. Ultimately, this effect must diminish as the opportunity cost of
additional growth in health spending rises—exacting a growing trade-off in the
forgone consumption of all other goods and services.
The authors thank Steve Heffler, John Poisal, and Sean Keehan for comments on an earlier draft. The views
expressed are those of the authors and do not represent the position of the Centers for Medicare and Medicaid
Services.
NOTES
1.
2.
3.
4.
5.
6.
7.
8.
Hartman M, Martin A, McDonnell P, Catlin A, et al. National health spending in 2007: slower drug spending contributes to lowest rate of overall growth since 1998. Health Aff (Millwood). 2009;28(1):246–61.
Ginsburg P. High and rising health care costs: demystifying U.S. health care spending. Princeton (NJ):
Robert Wood Johnson Foundation; 2008. Research Synthesis Report no. 18.
Fuchs V. Economics, values, and healthcare reform. Am Econ Rev. 1996;86(1):1–24.
Newhouse JP. Medical care costs: how much welfare loss? J Econ Perspect. 1992;6(3):3–21.
Newhouse JP. An iconoclastic view of cost containment. Health Aff (Millwood). 1993;12 Suppl:152–71.
For a concise summary of the findings of this literature, see Cohen AB, Hanft R. Technology and American
health care. Ann Arbor (MI): University of Michigan Press; 2004. p. 19–20.
See the technical appendix at http://content.healthaffairs.org/cgi/content/full/28/5/1276/DC1.
Manning W, et al. Health insurance and the demand for medical care: results from a randomized experi-
H E A L T H A F F A I R S ~ Vo l u m e 2 8 , N u m b e r 5
Downloaded from content.healthaffairs.org by Health Affairs on October 1, 2014
by guest
1283
O v e rv i e w
ment. Amer Econ Rev. 1987;77(3):251–77.
9. Newhouse J, Health Insurance Experiment Group. Free for all? Lessons from the RAND Health Insurance
Experiment. Cambridge (MA): Harvard University Press; 1993.
10. Getzen TE. Health care is an individual necessity and a national luxury: applying multilevel decision models to the analysis of health care expenditures. J Health Econ. 2000;19(2):259–70.
11. Finkelstein A. The aggregate effects of health insurance: evidence from the introduction of Medicare. QJ
Econ. 2007;122(1):1–37.
12. Newhouse J. Medical care expenditure: a cross-national survey. J Human Res. 1977;12(1):115–25.
13. Gerdtham U, Jönsson B. International comparisons of health expenditure. In: Culyer AJ, Newhouse JP, editors. Handbook of health economics. Amsterdam (Netherlands): North-Holland; 2000. p. 11–53.
14. Finkelstein A. Static and dynamic effects of health policy: evidence from the vaccine industry. Q J Econ.
2004;119(2):527–64.
15. Acemoglu D. Market size in innovation: theory and evidence from the pharmaceutical industry. Q J Econ.
2004;119(3):1049–90.
16. Acemoglu D, Finkelstein A. Income and health spending: evidence from oil price shocks. Cambridge
(MA): National Bureau of Economic Research; 2009. Working Paper no. 14744.
17. Weisbrod B. The health care quadrilemma: an essay on technological change, insurance, quality of care,
and cost containment. J Econ Lit. 1991;29(3):523–52.
18. Berndt ER, Cutler D, Frank RG. Medical care prices and output. In: Culyer AJ, Newhouse JP, editors.
Handbook of health economics. Amsterdam (Netherlands): Elsevier; 2000. p. 120–80.
19. Ho MS, Samuels JD. Appendix B: measuring industry output and intermediate inputs. In: An intertemporal general equilibrium model of U.S. health expenditures. Cambridge (MA): Dale Jorgenson Associates;
2006.
20. Baumgardner J. The interaction between forms of insurance contract and types of technical change in
medical care. RAND J Econ. 1991;22:36–53.
21. Garber A. Insurance and incentives for medical innovation. Cambridge (MA): National Bureau of Economic Research; 2006. Working Paper no. 12080. Incentives for innovation in drugs and devices need to
consider the world market, but the United States historically was a substantial proportion of that market.
22. Frank R, Zeckhauser R. Custom-made versus ready-to-wear treatments: behavioral propensities in physicians’ choices. J Health Econ. 2007;26(6):1101–27.
23. Sogaard J. Decomposition of the aggregate income elasticity in health care using panel data [Internet]. Copenhagen (Denmark): Danish Institute for Health Services Research; 2000 [cited 2009 Jun 24]. Available
from: http://www.nek.lu.se/ryde/luche99/Papers/econpapers/sogaard.pdf
24. This range reflects variation in sample across countries or time. Estimates for later time periods tend to be
on the lower end of this range.
25. Moise P. The technology-health expenditures link. In: A disease-based comparison of health systems. Paris
(France): Organization for Economic Cooperation and Development; 2003. p. 27–50.
26. Baumol WJ. Macroeconomics of unbalanced growth. Amer Econ Rev. 1967;57(3):415–26.
27. Phelps CE. The demand for reimbursement insurance. In: Rosett R, editor. The role of health insurance in
the health services sector. Cambridge (MA): National Bureau of Economic Research; 1976.
28. Centers for Medicare and Medicaid Services. Projections of national health expenditures: methodology
and model specification [Internet]. Baltimore (MD): CMS; 2009 [cited 2009 Jun 24]. Available from:
http://www.cms.hhs.gov/NationalHealthExpendData/downloads/projections-methodology.pdf
29. This assumes that the effect of insurance on spending does not increase.
1284
September/October 2009
Downloaded from content.healthaffairs.org by Health Affairs on October 1, 2014
by guest