Healthy Savings. Medical Technology and the Economic Burden of Disease

Anusuya Chatterjee, Jaque King, Sindhu Kubendran, and Ross DeVol
July 2014
HEALTHY SAVINGS
Medical Technology
and the Economic Burden of Disease

Anusuya Chatterjee, Jaque King,
Sindhu Kubendran, and Ross DeVol
July 2014
HEALTHY SAVINGS
Medical Technology
and the Economic Burden of Disease

ACKNOWLEDGMENTS
This project evolved from numerous discussions over the years with industry stakeholders, members of
the health policy community, and federal budget officials about the challenges of demonstrating medical
technology’s economic benefits relative to its costs. The study was made possible, in part, by support from
AdvaMed, the Advanced Medical Technology Association. The views expressed, and any errors or omissions,
are those of the authors and the Milken Institute. We are grateful to our colleagues at FasterCures, a center of
the Milken Institute, for the advice and expertise they provided. Additionally, we thank our research colleagues
Perry Wong and Robert Deuson for their helpful suggestions and support. Preliminary versions of this paper
were presented at the iHEA 9th World Congress on Health Economics, 2013, held in Sydney, Australia, and at
the 2014 AcademyHealth conference in San Diego. At both events, attendees made many suggestions that
enhanced the final form of this document. Lastly, we owe a debt of gratitude to our colleague and editor,
Edward Silver. He devoted many hours to significantly improving the quality and clarity of this report.
ABOUT THE MILKEN INSTITUTE
A nonprofit, nonpartisan economic think tank, the Milken Institute works to improve lives around the world
by advancing innovative economic and policy solutions that create jobs, widen access to capital, and enhance
health. We produce rigorous, independent economic research—and maximize its impact by convening
global leaders from the worlds of business, finance, government, and philanthropy. By fostering collaboration
between the public and private sectors, we transform great ideas into action.
©2014 Milken Institute
This work is made available under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License,
available at creativecommons.org/licenses/by-nc-nd/3.0/

CONTENTS
Executive Summary......................................................................................1
Overview....................................................................................................11
Technology and the Economic Burden of Disease: Historical Trends..........17
Economic Impact Projections and Medical Technology..............................39
Tax Revenue and Medical Technology........................................................61
Main Takeaways.........................................................................................63
Methodology.............................................................................................65
About the Authors....................................................................................103

1
EXECUTIVE SUMMARY
T
he debate continues within the health policy community on the proper balance between the
costs and benefits of medical technology. At a time of unprecedented change in health delivery
and incentive systems and persistent concern about the cost of care, this debate has significant
implications for public policy. Even with medical inflation running at a four-decade low—a condition that
might suggest pressures are dissipating—the controversy is only intensifying.
Assessments of the true cost and economic benefit of medical technology (in the form of devices
and diagnostics) have been hampered by the fact that direct treatment expenditures associated with
technology use can be readily measured, while indirect savings, for example avoiding emergency room
care and reducing hospital stays, are more difficult to capture.
Equally important, the economic benefits of reducing the burden of disease through better diagnosis,
prevention, treatment, and cures extend beyond the health system to GDP gains from increased labor force
participation and productivity. These gains are generated not only by patients, but by the rising participation
and productivity of their informal caregivers.Yet these dividends are rarely incorporated into the evaluation
of medical technologies.
In this study, we take a systematic approach to documenting the full costs and broader economic benefits
of investment in representative medical technologies used to address four prevalent causes of death and
disability: diabetes, heart disease, musculoskeletal disease, and colorectal cancer.1
The medical devices and
technologies analyzed for each of the four diseases examined are detailed in Table ES1.
Table ES1
Technology assessed in this study
DISEASE DEVICE OBJECTIVE
1) Diabetes i) Insulin infusion pumps Disease management
2) Heart disease i) Angioplasty Early detection/Disease management/Cure
ii) Pacemaker
iii) Electrocardiogram
iv) Left ventricular ultrasound
v) Chest x-ray
3) Musculoskeletal disease i) Joint replacement surgery Early detection/Disease management/Cure
ii) Bone scan (MRI)
4) Colorectal cancer i) Sigmoidoscopy Prevention/Early detection
ii) Colonoscopy
1. This analysis differs from the more common approach of estimating the number of quality-adjusted life years gained (QALY) from a technology
and comparing an estimate of the value of a QALY (conventionally $100,000 in the U.S.) to the cost of the technology. The estimates in this paper
define annual benefits in terms of actual dollars gained, either through a reduction in health costs enabled by the technology or increases in GDP.

2 Healthy Savings
We begin by estimating the annual net health system costs and additional impact on GDP of each
technology in 2010.2
·· We find that the net annual benefit from these technologies was $23.6 billion.
·· Federal income tax revenue increased by $7.2 billion due to improved labor market outcomes.
These estimates should be considered conservative because they do not account for reduced costs from
avoidance or amelioration of comorbidities, e.g., the impact of diabetes on heart and kidney disease.
Having assessed the most current net annual benefit of these technologies, we next construct three
alternative trajectories through 2035 for continued technological innovation for each of the four diseases
mentioned above. The first trajectory assumes reduced incentives to invest in improvement and adoption
and correspondingly reduced technological progress. The second trajectory assumes continuation of the
historical incentive level. The third assumes enhanced incentives.
·· The results demonstrate a cumulative $1.4 trillion gain (in 2010 dollars) over a 25-year period in the
“increased incentives” scenario relative to the persistence of “continued incentives.”
·· Conversely, the results indicate a cumulative $3.4 trillion loss (in 2010 dollars) over a 25-year period
in the “decreased incentives scenario” relative to “increased incentives.”
While this study does not examine specific policies that may affect incentives to invest in technology
development, it does make clear that such incentives have significant consequences for the economic
costs and benefits generated by the American health-care system. These should be considered in policy
development, especially at a time when the market forces and policies influencing health care are
changing dramatically.
The medical technologies studied generated economic returns that were substantially greater than their
costs. Policies that support enhanced investment in development, improvement, and diffusion of medical
technologies not only bring immense benefit to individual patients, but a brighter economic future for the
country as a whole. Conversely, reduced incentives will result in large net costs and, we believe, prove to be
pennywise and pound-foolish.
2. We used the annual average from 2008 through 2010 due to the small patient size in any one year and related high standard error of the sample.

3Executive Summary
Historical Experience
We find that these medical technologies are costly but provided substantial economic benefits in 2010
averaged across the population with the health condition that the technology targets.
Table ES2
Average annual savings per person affected associated with medical technology
2008-2010 ($)
TECHNOLOGY
TREATMENT
EXPENDITURES
INDIRECT IMPACT TOTAL
Insulin pump 607.7 5,278.0 5,885.8
Heart disease diagnostics and surgery -4,533.7 6,464.0 1,930.3
MRI and joint replacement surgery -3,887.3 28,405.2 24,517.9
Colonoscopy/sigmoidoscopy 8,840.7 141,524.2 150,364.9
Detection 903.5 96,398.5 97,302.0
Prevention 7,937.2 45,125.7 53,062.9
Sources: Medical Expenditure Panel Survey, National Health Interview Survey, Milken Institute.
·· Insulin pump use is associated with higher upfront costs than self-injection, but the net health system
expenditure per population reporting a condition (PRC) was $608 lower per pump user (see Table ES2).
Generally, pump users visited emergency rooms less frequently and were more able to avoid long-term
side effects such as amputations, kidney failure, or blindness. Additionally, GDP per person affected,
including informal caregivers, was $5,278 greater than the total for people who self-inject, due to higher
workforce participation and productivity. The net benefit per insulin dependent diabetic for pump use,
therefore, was $5,886: $608 in health cost savings and $5,278 in increased GDP.
·· Treatment expenditures per person reporting a condition for heart disease diagnostics and surgery
were $4,534 higher for technology users than non-users. However, higher survival rates and productivity
gains boosted real GDP per person affected by $6,464, resulting in a net economic impact of $1,930 per
person affected.
·· MRI and related joint replacement surgery expenditures were $3,887 higher than for other treatments per
PRC (person reporting musculoskeletal disease), but real GDP per person affected rose $28,405, contributing
to a net economic impact of $24,518. However, as with heart disease, there is an adverse selection bias in
the population represented by the data.The patients recommended for these procedures generally have
more advanced illness, which is more costly to treat. In these cases, less expensive alternatives were either
attempted and proved ineffective or the conditions had worsened before being diagnosed.
·· Treatment expenditures per PRC (person with colorectal cancer or per case prevented) undergoing
colonoscopy/sigmoidoscopy were $8,841 lower than those without screening due to the savings
from prevention and early detection. Additionally, GDP per person affected jumped $141,524
because screening also allows the removal of polyps before they develop into colorectal cancer.

4 Healthy Savings
Figure ES1
Economic effect associated with four medical technology areas
Average (2008-2010)
Net treatment
expenditures
Screening
expenditures
for healthy
population
Total net gain*
-51.6
-31.0
23.6
GDP gain
106.2
$ billions
-60
-40
-20
0
20
40
60
80
100
120
* Total net gain is the sum of treatment expenditures compared to non-users, screening of the healthy population, and the additional GDP
contribution of those receiving treatment and their caregivers.
Across these technologies, as shown in Figure ES1 above, overall treatment expenditures were $51.6 billion
higher than for non-user patients. Individuals who were screened but found to not have the disease added
another $31 billion to medical expenditures.That was concentrated in colorectal screening, with $17.4 billion.
The total includes the cost of screening people who expressed symptoms but turned out to be healthy and
those undergoing prescribed routine screening. The use of these technologies and treatments expanded
GDP by $106.2 billion (relative to non-use by the same population), which can be credited to higher survival
and labor force participation, less absenteeism, and greater productivity among patients and informal
caregivers. The net economic gain was $23.6 billion (synthesizing treatment expenditures for the four
technology areas compared to non-users, screening of the healthy population, and the additional GDP
contribution of those receiving treatment).
Alternative Futures
Investing in medical technology development is a high-risk endeavor, which largely stems from the sizable R&D
costs necessary as well as market and regulatory uncertainties.The environment for innovation and economic
returns will determine whether the industry can compete for capital effectively, which in turn will influence
the rate of technological progress and whether advances are broadly adopted. To evaluate the personal
and economic impact of incentives to innovate, we prepared three alternative scenarios through 2035:
·· Baseline (continued incentives)—the historical level of incentives that produced the 2008-2010 results,
·· Optimistic (increased incentives), and
·· Pessimistic (decreased incentives)

5Executive Summary
We do not tie the scenarios to explicit policy changes that might affect future innovations, such as medical
device taxes, reimbursement rules, or consequences of the Affordable Care Act. Nevertheless, these types
of policies were indirectly considered in constructing the various scenarios. If medical device taxes are
reduced or repealed, reimbursement or appropriate utilization rates increase, or the costs of regulatory
requirements associated with product development decline, the device industry is likely to invest more in
innovation and follow the increased incentives projection. Similarly, factors that erode future profitability
make the decreased incentives scenario more likely.
Our approach to projecting treatment expenditures under these alternative paths involves comparing
projected outcomes resulting from different assumptions about the improvement and diffusion of disease-
specific technology.
To generate these results, we used decision trees that include disease-specific Markov models.These models
identify disease stages and the probability of transitioning from one stage to another. The different values
for the three scenarios by disease are contingent on assumptions of the potential for technological progress
and its impact on individuals’progression through the disease states. These probabilities drive the differing
cost estimates for each scenario and were developed from our review of the literature and discussions
with specialists. While our decision trees differ by disease, all have the same basic structure beginning with
three health states: well, sick, or dead (of any cause). A probability is associated with each state and any
subsequent branch of these states.
For each projected year, the number of people reporting the relevant condition for each health state is
computed.The aging of the population and rising obesity rates are the primary drivers of increasing chronic
disease rates. The per-person cost of each condition and each health state is derived from the Medical
Expenditure Panel Survey, compiled by the U.S. Agency for Healthcare Research and Quality, and the overall
costs of the disease are calculated. The difference in economic impact among the scenarios demonstrates
the benefits and losses associated with investing in medical technology innovation.
As mentioned earlier, the estimates are conservative for not considering savings from avoiding or
ameliorating comorbidities. In addition, the technological improvements assumed in the model are
incremental and do not consider potentially transformative technologies that could produce a greater
impact on treatment economics and GDP. Hypothetical examples might include an artificial pancreas for
type I diabetes, an inexpensive blood screening test for colorectal cancer, or tissue regeneration techniques
to forestall or delay knee and hip replacements.
Aggregate savings stem from the increased incentives
scenario relative to continued incentives. By 2035, the savings
are projected to reach $217 billion. Decreased incentives
result in dissavings of $470 billion.

6 Healthy Savings
Figure ES2
Aggregate savings from medical technology
$ billions $ billions
2010
Savings from increased incentives
Compared to continued incentives
Losses from decreased incentives
Compared to continued incentives
2010 2015 2020 2025 203520302015 2020 2025 2030 2035
0
40
80
120
160
200
240
-500
-400
-300
-200
-100
0
28.21
0.00
0.00
-18.82
-59.81
-131.84
-256.72
-469.89
70.51
112.42
160.08
217.37
Aggregate savings, as seen in Figure ES2, stem from the increased incentives scenario compared to
continued incentives. By 2035, the savings reach $217 billion. Conversely, decreased incentives result in
dissavings of $470 billion.
Applying the model to each disease state produced the following results:
·· Greater use of insulin pumps among the insulin-dependent PRC and improvements in efficacy will pare
treatment costs and expand economic activity in the increased incentives projection relative to the
other two scenarios. Increased incentives would reduce treatment expenditures by $19.6 billion
and expand GDP by $205.8 billion over 25 years in 2010 dollars compared to the continued incentives
scenario. Similarly, increased incentives would decrease treatment expenditures $28.9 billion and boost
GDP by $297.6 billion compared to decreased incentives.
·· Heart disease diagnostics and surgical procedures, assuming expanded use and efficacy, would create
substantial health and economic gains through 2035. Treatment costs are $35.4 billion lower and
GDP grows $773.7 billion under increased incentives relative to the continued incentives scenario.
Treatment costs are $224.9 billion lower and GDP jumps $2.1 trillion with increased incentives relative
to the decreased incentives scenario.
·· MRI and related joint replacement surgery are projected to be increasingly common due to rising obesity
and age-related disease. Treatment costs are $30.6 billion lower and GDP increases $250.4 billion in the
increased incentives scenario relative to continued incentives. Treatment costs are $62.2 billion lower
and GDP rises $527.7 billion in the increased incentives scenario relative to decreased incentives.
·· The health and economic benefits of colonoscopy/sigmoidoscopy will be even greater in the future
than today. Treatment costs over the 25 years are $27.3 billion less in the increased incentives scenario
compared to continued incentives, and GDP grows by $150.8 billion in 2010 dollars. Treatment costs
are $44.6 billion lower in the increased incentives scenario, and GDP elevates by $245.3 billion,
compared to decreased incentives.

7Executive Summary
Figure ES3
Effect of increased medical technology incentives
Compared to continued incentives, 2010-2035 (2010 dollars)
$ billions
-200
0
200
400
600
800
1,000
1,200
1,400
1,600
Net treatment
expenditures
(savings)
GDP gain Total net gain*
113.0
1,359.81,380.8
Screening
expenditures
for healthy
population
-134.0
* Total net gain is the sum of treatment expenditures compared to non-users, screening of the healthy population, and the additional GDP
contribution of those receiving treatment and their caregivers.
As highlighted in Figure ES3, from 2010 to 2035, the combined health and economic benefits of the
increased incentives scenario far outstrips those of continued incentives. In our four areas, $113.0 billion
is saved in treatment costs and GDP rises by $1.38 trillion due to more people working and doing so more
productively. Subtracting the higher costs of screening healthy people, which amounts to $134 billion,
the net result is a gain of $1.36 trillion in 2010 dollars.
Similarly, from 2010 to 2035, the combined health and economic benefits generated by the increased
incentives scenario surpasses those of decreased incentives by an even wider margin, with treatment cost
savings of $360.5 billion and GDP gains of $3.2 trillion. Subtracting the higher costs of screening healthy
people, which amounts to $197.9 billion, the net result is a $3.4 trillion boost in 2010 dollars.
The Broader Picture
Along with measuring their impact on health costs, an evaluation of new medical technologies should
incorporate the broader benefits of preventing premature death and improving the capacity of patients and
caregivers to contribute to economic growth. Calculating the economic value generated by these technologies
is a challenge, but applying a consistent, balanced methodology can yield useful and relevant results.
Our projections demonstrate the economic value of raising incentives to innovate in representative
technologies used to diagnose and treat these diseases—a finding we believe likely applies to other
technologies and ailments as well. Conversely, if the costs associated with regulatory and market conditions
are higher in the United States than those of other countries, fewer medical innovations will emerge within
U.S. borders. Better incentives would help spur research breakthroughs, expand the size and productivity
of the workforce, create more high-paying jobs in devices and diagnostics, and contribute to the economy
across the board—a healthy combination.

8 Healthy Savings
Sedentary lifestyles
Greater need
for technology
Increasing number
of people affected
by disease
Aging population
Rise in obesity
Unhealthy diet
Health care system
Increasedcosts
duetoscreening
thehealthy
population
Changesin
treatment
expenditures
The overall economy
Increasedtax
revenuefromboth
patientsandcaregivers
Betterlabormarket
outcomes/
increasedGDP
Productivityfor
bothpatientsand
caregiversrises
Improvedsurvival
expandsworkforce
Disease
incidence
prevented
Disease
detection
improved
Number of cases
Survivalboosted
forthose
withdisease
PREVENTION EARLY DETECTION DISEASE MANAGEMENT
MED-TECH ADDRESSES A GROWING NEED
MED-TECH FACILITATES
MED-TECH’S EFFECTS
THE SCOPE OF MEDICAL TECHNOLOGY

9Executive Summary
The insulin pump, an innovative technology for diabetes patients, improves
management of the disease. Among users and caregivers, the average annual
savings per person was $5,886 compared to non-users of the device. Most of
that benefit came from savings to GDP, as patients and their caregivers missed
fewer workdays and were more productive. Expanding innovation in diabetes
management would increase aggregate savings by $225.4 billion.
Angioplasty, an innovative procedure used to treat heart disease, is likely to
generate substantial savings in the future. This technology, in combination with
electrocardiogram, echocardiogram, chest x-ray, and pacemakers, saved $1,930 per
person affected annually compared to those who did not use technology. Increased
incentives, which spur technology innovation and expand use, would lead to long-
term savings of $809 billion compared to the continued incentives scenario.
Joint replacement can elevate quality of life for musculoskeletal disease
patients and even cure disease. This technology, in conjunction with MRI
screening, saved $24,518 annually among users and caregivers compared to
non-users. Increasing the incentives for innovation in musculoskeletal disease
technology would save $281.1 billion compared to continuing current incentives.
Colonoscopy and sigmoidoscopy detect colorectal cancer, and colonoscopy
can prevent the ailment through the removal of polyps. These technologies
led to an average annual savings per person affected of $97,302 compared to
unscreened patients. In addition, screening that prevented colorectal cancer
saved $53,063 per case. Aggregate savings associated with innovation in this
field would amount to $178.2 billion due to improved detection and prevention.
Average annual savings are based on 2008-2010 data. Projections represent aggregate savings over 25 years in 2010 dollars.
A PHYSICAL AND ECONOMIC PAYBACK

11
OVERVIEW
A
s America ages and sedentary lifestyles and unhealthy diets become more common, the nation
is likely to suffer a sharp rise in the prevalence of chronic disease during the 21st century. As that
future unfolds, technology, in the form of advanced diagnostic and therapeutic devices, can answer
the need for early detection and more effective management of illness. Cost is a crucial element of the value
proposition for such technology, however, along with the benefits it brings. Deepening our knowledge of
how these tools affect both treatment expense and the link between health and productivity would aid
decision-making around developing these technologies and provide a more informed basis for public policy.
A review of the research on this topic brings to light fragmented and sometimes contrasting results.
While much of the literature seems to demonstrate that successive generations of medical technology
have prevented countless deaths and improved the quality of life for millions more, other researchers
have questioned whether the overall benefits of these technical advances—early-diagnostic tests, devices,
and the procedures they enable—outweigh the costs.
One group believes that medical technologies have pushed costs up because of overutilization and
unnecessary, expensive testing and procedures. However, others point to the benefits of early detection,
such as higher survival rates and disease prevention; reduced use of cost-intensive therapeutic settings,
including fewer inpatient hospital days and emergency room visits; and economic growth through
increased productivity.
Accordingly, this report undertakes a comprehensive, quantitative documentation of medical technology’s
impact on the economic burden of disease. It estimates changes (if any) in treatment expenditures and
workforce productivity associated with these tools. Further, it projects how future innovation in this sector
would affect the health-care system and the larger economy.
The utility and value of such investments are considered by examining innovations pertaining to four prevalent
causes of disability and death: heart disease, diabetes, colorectal cancer, and musculoskeletal diseases.
This report uses the term“medical technology”to describe medical devices often used for therapeutic and
diagnostic purposes for the diseases mentioned above.Therapeutic devices such as insulin pumps and pacemakers
treat diseases or disorders. On the other hand, diagnostic devices such as colonoscopy and magnetic resonance
imaging (MRI) equipment are used to identify a patient’s health status before, during, or after a treatment.
These devices are typically developed through a collaboration between clinician and manufacturer in an
effort to respond to an unmet need. Often, a manufacturer will modify an existing device to create a new
generation of the product intended to improve patient care outcomes. As an example, the technology
behind cardiac resynchronization therapy with defibrillator (CRT-D) has undergone several cycles of
improvement. A device that sends electrical impulses to the heart and can detect and treat irregular heart
rhythms, the CRT-D is also a tiny computer. One of the implanted wires transfers signals from the heart to
external devices that aid doctors in prescribing the appropriate treatment. Now it features wireless remote
monitoring, which enables the collection of diagnostics on a patient’s heart disease in real time.
Over the long term, better monitoring and detection of disease reduces the need for expensive forms of
care (such as emergency room visits) and raises the productivity of working people. Manufacturers and

12 Healthy Savings
clinicians play key roles in innovating and updating devices to serve the needs of patients. However, before a
device becomes available to the public, it must be approved by a regulatory agency. In the United States,
the Food and Drug Administration must review and approve new devices and device modifications.
This study considers therapeutic and diagnostic devices that are widely used and have substantially
affected the lives of patients and their caregivers as well as the overall U.S. economy. We also note that the
effectiveness of a medical device largely depends on the type and intensity of the disease and is influenced
by the skills of clinicians, the complications associated with a procedure, and patient compliance.
Table 1
Technology assessed in this study
DISEASE DEVICE OBJECTIVE
1) Diabetes i) Insulin infusion pumps Disease management
2) Heart disease i) Angioplasty Early detection/Disease management/Cure
ii) Pacemaker
iii) Electrocardiogram
iv) Left ventricular ultrasound
v) Chest x-ray
3) Musculoskeletal disease i) Joint replacement surgery Early detection/Disease management/Cure
ii) Bone scan (MRI)
4) Colorectal cancer i) Sigmoidoscopy Prevention/Early detection
ii) Colonoscopy
In this report, therefore,“medical technology”will refer to the devices listed above in Table 1. Due to the lack
of sufficient data to differentiate the effects of each device, we provide evidence of the combined effect of
technology on each disease. For instance, regarding musculoskeletal illness, we examine the effect of having
an MRI as a diagnostic tool and/or joint replacement surgery as a means of treatment. The assumption is that
this surgery is usually preceded by an MRI and often followed by one. Hence, separating the effects of the MRI
as a diagnostic from the surgery is not realistic for patients undergoing surgery. At the same time, however,
it is necessary to include patients whose condition was detected by an MRI at an early stage and less invasive
treatment was prescribed.
Many would ask why X-ray technology is not included. X-rays are routinely used to examine musculoskeletal
disease, but we chose to consider MRI because that technology offers potentially more accurate results and
faster diagnosis, producing a greater impact on the cost of care and labor market outcomes.
This report starts by assessing the historical data on how devices affect the economic burden of the
diseases studied. Further, we project the effects of advancing technology on the future economic burden
of each disease. Three scenarios are posed to quantify potential savings generated by incentives for the
future availability and advancement of these technologies: a baseline “continued incentives” scenario,
an optimistic “increased incentives” scenario, and a pessimistic “reduced incentives” scenario. The study
estimates these alternative pathways through 2035.

13Overview
However, we do not explicitly incorporate policy changes that might affect innovations in the future,
such as medical device taxes, reimbursement rules, or any consequence of the Affordable Care Act.
These assumptions are implicit in various incentive scenarios. If medical device taxes are reduced or repealed,
for example, or reimbursement levels increase, the device industry is likely to invest more in innovation
and follow the increased incentives projection. Similarly, factors that erode future profitability make the
decreased incentives scenario more likely.
The data demonstrate that the use of medical technology brings considerable economic benefits. They are
expressed in the aggregate savings in treatment expenditures and prevention as well as the reduction of
“indirect impact”through larger contributions to the economy. Though treatment expenditures are relatively
straightforward, the concept of indirect impact is more difficult to grasp, though it is essential to accounting
for the effects we are investigating.
A disease can substantially influence labor market outcomes. Employees suffering from ailments miss
workdays, a situation known as absenteeism, or perform far below their potential, which is called
presenteeism. Informal caregivers also may experience absenteeism and presenteeism. As a result,
businesses suffer and the productivity of the entire economy declines, along with GDP. Medical devices
might diminish this indirect impact (measured in terms of foregone GDP) through better disease
management, prevention, or cure. For example, joint replacement surgery can relieve pain, dramatically
reduce sick days, and raise productivity. This technology often improves the chances of curing a patient’s
condition, can extend his or her survival and can boost the economy through expanded workforce
participation and stronger performance on the job.
When we discuss the economic
burden associated with medical
device use, we can’t ignore the
effect of screening the non-patient
population. Although it is widely
acknowledged that screening aids
early detection, the technology
is often considered overused,
considering that the majority of
people to which it is applied will
not have the disease. Increasing
the rate of screening raises the
health-care system’s outlays.
This must be considered when
examining the costs of a medical
technology.
So to capture the effect of medical device use on the health system and the broader economy, we define
the economic burden as the aggregate of disease treatment expenditures, indirect impact for individuals
and informal caregivers (measured by lost GDP), and diagnostic spending for non-patient populations.
Table 2 shows that for 2008 through 2010, the average annual economic burden associated with insulin
pump use was $3.2 billion. Similar values for heart disease and musculoskeletal disease technology were
$102.8 billion and $44.9 billion, respectively.
Technology-related gains associated
with heart and musculoskeletal
disease were $1,930 and $24,518
per person affected, respectively.
Technology did not reduce cost
of care, but better quality of life
and survival rates contributed to
economic gains generated by higher
workplace productivity.

14 Healthy Savings
Colorectal cancer screening can affect the health-care system and GDP through early detection of the disease.
However, an important source of value created by such screening is prevention through the removal of potentially
cancerous polyps.We estimate the economic burden associated with colorectal cancer screening at $22.5 billion.
It would have been much more, but the burden was offset by $12.2 billion in gains linked to prevention.
Table 2
Average annual economic burden associated with medical technology
2008-2010 ($ millions)
TECHNOLOGY
TREATMENT
EXPENDITURES
INDIRECT
IMPACT
HEALTHY
SCREENING
TOTAL
Insulin pump 1,223 1,993 - 3,216
Heart disease diagnostics and surgery 62,604 33,685 6,522 102,812
MRI and joint replacement surgery 23,103 13,473 8,302 44,878
Colonoscopy/sigmoidoscopy 216 4,834 17,445 22,495
Detection 4,711 12,557 17,445 34,713
Prevention -4,495 -7,723 - -12,218
Although the economic burden summarizes the aggregate contributions of each device studied, the business
rationale behind their use is justified by measuring the savings per person affected.“Person affected”includes
patients, or the population reporting a condition (PRC), when assessing treatment expenditures. A part of
this group is employed, which we refer to as the employed population reporting a condition (EPRC), and they
affect the economy through foregone GDP. In addition, employed caregivers by condition (ECC) for these
patients affect the labor market and in turn the economy.“Person affected,”therefore, refers to one or all of
the above groups as appropriate for the analysis.
For insulin pump users, Table 3 shows that savings to the health-care system and the economy was
equivalent to $5,886 annually per person affected, for 2008 through 2010. The majority of savings came
from the increased economic contribution of $5,278 per person affected.
Technology-related gains associated with heart and musculoskeletal disease were $1,930 and $24,518,
respectively. For both of these diseases, technology did not reduce the cost of medical care. However,
improved quality of life and higher survival rates contributed to significant economic gains generated by
higher workplace productivity. Hence, the use of devices in these disease categories can be justified by
improved labor market outcomes. As mentioned earlier, colorectal cancer screening not only facilitates
early detection but has proved beneficial for prevention. Our estimates show that the annual per-person
savings from such screening were $150,365, with $97,302 from early detection and $53,063 from prevention.

15Overview
Table 3
Average annual savings per person affected associated with medical technology
2008-2010 ($)
TECHNOLOGY
TREATMENT
EXPENDITURES
INDIRECT
IMPACT
TOTAL
Insulin pump 607.7 5,278.0 5,885.8
Heart disease diagnostics and surgery -4,533.7 6,464.0 1,930.3
MRI and joint replacement surgery -3,887.3 28,405.2 24,517.9
Detection 903.5 96,398.5 97,302.0
Prevention 7,937.2 45,125.7 53,062.9
To the extent that medical technology enables employees to work longer and more productively,
they contribute more in income and other taxes. Our study estimates the amount of federal income tax
revenue added or lost due to changes in labor market outcomes.Technology associated with our examined
diseases could have increased tax revenue by an annual average of $7.2 billion. An annual increase of
$34.9 million in tax revenue could have been generated by insulin pump use. Technology use for heart
disease could have generated an additional $1.5 billion in tax revenue, and $3.8 billion by technology that
addresses musculoskeletal disease. Colorectal cancer screening has the potential to expand tax revenue by
$1.8 billion. Of this, $1.3 billion stems from early detection.
Table 4
Federal tax revenue associated with medical technology
Compared to non-users
2008-2010 ($ millions)
TECHNOLOGY AVERAGE (2008-2010)
Insulin pump 34.9
Heart disease diagnostics and surgery 1,474.1
MRI and joint replacement surgery 3,798.1
Colonoscopy/sigmoidoscopy 1,844.2
Detection 1,318.9
Prevention 525.3

16 Healthy Savings
Along with the evidence of considerable savings produced by the use of medical devices, there is concern
about whether future innovations will be worth the investments required. To investigate, we calculated
the projected economic impact for each disease, as seen in the table below. We consider three future
scenarios that simulate the growth rates of technology innovation. Based on these, we conclude that more
innovation in this field might result in larger numbers of patients (or PRC) and thereby increase overall
treatment expenditures. However, it might pare back the average cost because better disease management
reduces expensive site of service visits and creates value in the labor market.
Hence, expanding innovation in the management of diabetes will increase aggregate economic savings
by $225.4 billion in 2010 dollars over 25 years. Similarly, aggregate savings from accelerating device
innovations for heart and musculoskeletal ailments will raise the economic contribution to $809.1 billion
and $281.1 billion, respectively. Aggregate savings associated with colorectal cancer are $178.2 billion due
to early detection and prevention. By the same logic, less investment in medical technology might have the
opposite effects.
Table 5
Projected economic impact by disease
2010-2035 ($ billions*)
ABSOLUTE DIFFERENCE
CONTINUED
INCENTIVES
INCREASED
INCENTIVES
DECREASED
INCENTIVES
CONTINUED-
INCREASED
CONTINUED-
DECREASED
Diabetes 12,342.6 12,117.2 12,443.7 225.4 -101.1
Heart disease 7,737.3 6,928.2 9,288.6 809.1 -1,551.3
Musculoskeletal
disease
24,673.5 24,392.4 24,982.3 281.1 -308.8
Colorectal cancer 2,005.2 1,885.5 2,072.6 119.7 -67.4
Colorectal cancer
prevented
-452.0 -510.5 -407.7 58.5 -44.2
* In 2010 dollars.

17
TECHNOLOGY AND THE ECONOMIC
BURDEN OF DISEASE: HISTORICAL TRENDS
T
he influence of medical devices on the economic burden of disease is illuminated by studying historical
trends.This report uses a cost-of-illness approach to examine trends from 2005 to 2010.“Economic
burden”is defined as the aggregate of direct treatment expenditures, indirect economic impact (in
terms of foregone gross domestic product), and costs for screening the healthy population.The benefit or loss
of using technology is measured as the difference between the economic impact of using the technology to
treat a disease and the economic effect of not doing so for the same purpose.
We calculated disease-related treatment expenditures and number of patients, which we refer to throughout as
the population reporting a condition (PRC) from the Medical Expenditure Panel Survey (MEPS).That information
is collected by the Agency for Healthcare Research and Quality (AHRQ), a unit of the U.S. Department of Health
and Human Services. MEPS is a nationally representative sample of the noninstitutionalized civilian population
with data on the provision of health services, site of service, frequency, and associated payments.
The MEPS database uses medical condition codes and ICD-9 codes to indicate the conditions for which each
patient is treated. Disease-related expenditures were calculated as aggregate expenditures of visits associated
with the relevant condition codes. Expenditures rather than charges were used to ensure that all costs levied
on the health-care system were included. For example, expenditures were calculated for diabetes-related
visits to offices, outpatient, inpatient, emergency room, and home health settings, and prescriptions for
each year. The same was calculated for all other diseases. PRC was the number of unique patients with visits
associated with a condition at any site of service.
Chronic diseases such as those assessed in this report are often accompanied by other ailments such as heart
failure, renal diseases, blindness, etc. However, due to a lack of available data and the risk of double counting,
our estimates did not take into account the economic impact associated with such comorbidities. In our
assessment of per-PRC expenditures, patients are identified as technology users if they have a technology-
related treatment expenditure in that calendar year. Therefore, these calculations do not capture any change
in cost of care in the years following that use. If a person uses less care due to improved symptoms after joint-
replacement therapy, this would not be captured in our analysis. Our approach can be seen as conservative.
The cost of screening for the non-patient population has a major impact on the health-care system, which must
be included in estimating the overall economic burden tied to disease-specific medical technology. For most
diseases, we used MEPS, the Healthcare Cost and Utilization Project (HCUP), scientific literature, and market
research to acquire information on the number of healthy people screened and the average (unit) cost,
enabling us to estimate the total cost of such screening.
Our calculation of indirect impact measures labor market outcomes related to work loss and productivity.
It represents the combination of absenteeism, or lost workdays due to disease, and presenteeism,
or underperformance at work for the same reason, and is quantified in terms of lost employee output,
or foregone GDP. We incorporate the absenteeism and presenteeism of both patients and their informal
caregivers to capture the total indirect impact of a disease.

18 Healthy Savings
The main source for lost workdays data associated with a disease was the National Health Interview Survey
(NHIS). The survey asks a nationally representative sample health-related questions regarding medical
conditions, employment, treatment, and cancer screening. The employed population reporting a condition
(EPRC) and lost workdays were calculated from a survey question about whether participants had missed
work due to illness or injury. A GDP-based approach was used to estimate the value of lost workdays,
or absenteeism. Then presenteeism was estimated using disease-specific presenteeism-to-absenteeism
ratios from a study by Goetzel et al.3
The number of employed caregivers by condition (ECC) and caregiver lost workdays were estimated using
studies from the National Alliance for Caregiving and AARP.4
Using a similar GDP-based approach, the value
of caregiver absenteeism was calculated. Further, caregiver presenteeism was estimated using information
from a study by Levy and indexed to employed patients’ (EPRC) presenteeism.5
Once we estimated the indirect impact of the overall disease, it was necessary to estimate the indirect
impact associated with the use of medical technology. In many cases, such technology can lower the
indirect impact associated with a disease because of better labor market outcomes. For example, a device
that eases arthritis pain can improve an employee’s job performance. The difference in the indirect impact
between device use and no use is the value added to or subtracted from the GDP.
DIABETES
Examining historical trends, we find that:
·· The average annual U.S. (2008-2010)6
economic
impact associated with insulin pump use was
$3.2 billion. To break that down, direct treatment
and disease management expenditures were
$1.2 billion and lost GDP amounted to $2 billion.
(See Summary Chart: Diabetes.)
·· Due to better disease management, the average
annual (2008-2010) savings per person affected was
$5,886 compared to insulin-dependent patients who
did not use pumps.This is due to the smaller economic
impact associated with pump use compared to other
modes of insulin delivery. The greatest portion of this
benefit stems from an economic gain of $5,278 per
person affected amid rising productivity and fewer
lost workdays.
3. Goetzel et al. “Health, Absence, Disability, and Presenteeism Cost Estimates of Certain Physical and Mental Health Conditions Affecting U.S.
Employers,” Journal of Occupational and Environmental Medicine 46, (2004).
4. National Alliance for Caregiving (NAC) and AARP. “Caregiving in the U.S.,” 2009.
5. David Levy. “Presenteeism: A Method for Assessing the Extent of Family Caregivers in the Workplace” (American Association for Caregiver
Education, 2003). See also David Levy. “Presenteeism: A Method for Assessing the Extent of Family Caregivers in the Workplace and Their
Financial Impact” (American Association for Caregiver Education, 2007).
6. Data was calculated annually for the period 2005-2010. Average annual economic impacts, the sum of treatment expenditures and indirect
impact, are calculated for 2008-2010.
»» Diabetes affects 25.8 million Americans
(more than 8 percent of population).
»» 7 million of this 25.8 million are
undiagnosed.
»» 5.4 million Americans are insulin
dependent (including all type 1 diabetics).
»» Deaths from diabetes-related disorders:
• 28 percent caused by cerebrovascular
disease
• 55 percent caused by renal failure
Sources: Centers for Disease Control and
Prevention, American Diabetes Association,
Diabetes/Metabolism Research and Reviews.

19Technology and the Economic Burden of Disease: Historical Trends
Diabetes is a chronic disease involving the
loss of sensitivity to the insulin hormone and/
or loss of the pancreas’ ability to produce it.
There are two types of the disease. Type 1
is auto-immune, always insulin dependent,
and generally occurs at an early age. Type 2
is more affected by risk factors such as diet
and exercise, has an older age of onset, and is
insulin-dependent primarily in severe cases.
Regular dosing and monitoring is necessary for
insulin-dependent patients.
Traditionally, injection is the mode of
administering insulin. However, pumps are
now gradually supplanting them. Although the
MEPS survey collects information about the
number of insulin-using diabetics, it does not
distinguish by mode of administration. Bode et
al. reported historical data on the number of
insulin pump users, and we used interpolation
to determine values for missing years. We
combined this with data on insulin users from
the Centers for Disease Control and Prevention
(CDC) to estimate the historical percentages of pump users, which steadily increased from 2000 to 2010.7
The rise in pump use over time may be explained by technology improvements that increased accuracy
and ease of use combined with research confirming pumps’ efficacy in disease management.
Figure 1
Proportion of insulin dependent diabetes patients using a pump
Percent
0
1
2
3
4
5
6
7
8
2000
2.7
3.1
3.5
4.0
4.4
4.8
5.3
5.7
6.2
6.6
7.1
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
7. Bruce W. Bode et al., “Diabetes management in the new millennium using insulin pump therapy,” Diabetes/Metabolism Research and Reviews 18,
Suppl. 1 (2002), pp. S14-S20.
»» Insulin pumps (also known as continuous
subcutaneous insulin infusion, or CSII therapy) deliver
the hormone to the bloodstream through a catheter
placed under the skin. The device is connected to
a pump (about the size of a pager) programmed to
deliver a specific amount continuously, which can be
monitored by the patient.
»» At present, fewer than 30 percent of type 1 patients
and 1 percent of type 2 patients are using insulin
pumps. This technology is likely to see further
adoption because of its ease of use and improved
ability to regulate blood sugar.

20 Healthy Savings
PRCs associated with sites of service were determined separately from overall insulin pump use to enable
us to estimate pump-related expenditures. Initially, the proportion of pump users to total insulin users was
applied across all sites of service to get a base number for pump-user PRC. However, we know that pump
use affects health outcomes and therefore changes health-care utilization patterns. We accounted for this
through additional percentage reductions in the pump user PRC. Scuffham and Carr8
report that insulin
pump use is associated with fewer hypoglycemic events (inpatient hospital stays and/or emergency room
visits). As a result, PRC for inpatient admission falls by an estimated 43 percent for insulin pump users and
PRC for ER visits 53 percent from the base number. It is logical to assume that all insulin pump users are
included in the prescription-related PRC. Since some insulin users may not incur Rx expenditures over the
course of a year, the use of the base number provides an upper bound.
There was no available data relating to changes in office-based and outpatient care. However, diabetic
patients need to regularly visit their clinician regardless of their status. Therefore, pump use would not
reduce office-based and outpatient services as much as it would reduce ER visits or inpatient admissions.
We assumed a 35-percent reduction, smaller than those for ER and inpatient care. Using the above research,
related expenditures were estimated using similar methodology but with different values. Aggregating
expenditures by site of service enabled us to estimate the total annual treatment expenditures for diabetes.
To be comprehensive, we also wanted to quantify the indirect impact of the disease. Diabetes is a disease
that can have dramatic adverse effects on labor market outcomes in terms of lower participation and
productivity losses. After calculating the indirect impact associated with overall diabetes and also insulin
dependent patients using previously described methods, the challenge was to estimate the indirect impact
associated with using insulin pumps.
The proportion of insulin pump PRC was used to calculate associated EPRC. We assumed that the better
disease management tied to pump use would improve labor market outcomes and reduce absenteeism
and presenteeism. To estimate the associated reduction in absenteeism, data from a study by Scuffham
and Carr9
(demonstrating that pump use reduces hypoglycemic events 13 percent) was used to adjust for
lost workdays. We acknowledge that hypoglycemic events are not the only drivers of diabetes-related
labor market outcomes. However, due to lack of consistent data on other types of diabetic complications,
we referred only to the hypoglycemic events.
One advantage of the reduction in diabetic complications is that patients feel less anxious and their quality
of life improves, reducing presenteeism as well. Research shows that the quality of life for those who inject
insulin is 5.3 percent worse than those using pumps.10
We assumed that pump users’presenteeism was
5.3 percent less than that of diabetes patients overall.
Overall annual treatment expenditures for diabetes rose from $34.2 billion in 2005 to $51.2 billion in 2010,
and the proportion of insulin-dependent diabetics increased from 20.3 to 24.4 percent. Those who are
insulin dependent represented 46.4 percent of diabetes-related expenditures and 18.7 percent of the
indirect impact in 2010, a significant portion of the total economic impact.
8. P. Scuffham and L. Carr. “The Cost-Effectiveness of Continuous Subcutaneous Insulin Infusion Compared with Multiple Daily Injections for the
Management of Diabetes,” Diabetes Medicine 7 (2003) pp. 586-93.
9. Ibid.
10. Ibid.

Total treatment expenditures associated with insulin pump users are also on the rise, probably due to increased
use. Average per-PRC savings to the health-care system due to using insulin pumps was $607.7 between 2008
and 2010. This per-PRC savings presents an economic rationale for their use.
We estimate that in 2010, the indirect impact for pump users and their caregivers was $2.3 billion, a small
portion of the $208.8 billion that represents the total indirect impact of diabetes. Using insulin pumps
increased GDP per person affected by $4,772 compared to other modes of insulin administration.
Some of these savings in treatment expenditures and indirect impact could be explained from earlier
research in this field. Studies have found that insulin pump therapy has resulted in at least equivalent,
if not lower, levels of HbA1c, or hemoglobin A1c.11
Better disease management leads to maintaining those
levels below 7 percent, a common target for diabetic patients.12
Indeed, the use of insulin pumps lowers
HbA1c levels 1.2 percent compared to multiple daily injections.13
The devices more closely replicate the
insulin production patterns of the pancreas, cutting the risk of diabetic complications such as nocturnal
hypoglycemia (low blood sugar) and early-morning spikes in blood sugar.14
This reduces the need
for expensive inpatient and emergency room care as well as lost workdays caused by such events.
Because pumps require less maintenance, workplace productivity is also improved.
Treatment expenditure data from 2010 supports the idea that insulin pumps better manage disease
and generate savings. In 2010, office-based and outpatient expenditures per PRC were approximately
50 percent lower for insulin pump users, indicating that non-pump delivery methods may require closer
clinician management. Poor blood sugar management is associated with a number of harmful effects,
including nephropathy, neuropathy, and retinopathy, which may require surgical management after
progression and increase hospital admissions. Insulin pump use appeared to reduce the probability of
admission, and in fact, inpatient expenditures per PRC for pump users were 60 percent lower than for non-
pump users. Further, the 80 percent reduction in per-PRC emergency room expenditures for pump users
may be attributed to a lower likelihood of hypoglycemic and hyperglycemic events.
This easing of the progression of diabetes-related complications can also explain the 50 percent reduction
in average home health expenditures associated with pump use, which can facilitate tight glucose control
and ultimately prevent complications that require greater nursing care. The only site of service that was
more costly for insulin pump users was prescription-related expenditures; the 40 percent increase in
spending in that category can be attributed to the price and maintenance of the device itself. So, even
though out-of-pocket prescription expenses rise with the use of insulin pumps compared to injections, it is
justified by the savings from fewer visits to expensive sites of service.
11. Bruce W. Bode, “Insulin Pump Use in Type 2 Diabetes,” Diabetes Technology & Therapeutics 12, Suppl. 1 (2010) S17-S21.
12. Ibid.
13. Meaghan St. Charles et al., “A Cost-Effectiveness Analysis of Continuous Subcutaneous Insulin Injection versus Multiple Daily Injections in Type 1
Diabetes Patients: A Third-Party U.S. Payer Perspective.”
14. Bruce W. Bode, “Insulin Pump Use in Type 2 Diabetes.”
Using insulin pumps increased GDP per person affected by
$4,772 compared to other modes of insulin administration.

22 Healthy Savings
SUMMARY CHART: DIABETES
$ thousands
2005 2006 2007 2008 2009 2010
10
15
20
25
30
35
40
45
22.3
27.5
29.0
35.6
27.7
39.0
24.9
30.6
22.8
29.4
24.9
30.4
Insulin pump Insulin, no pump
Economic impact of insulin dependent diabetes, 2005-2010*
Per person affected
* Includes treatment expenditures and indirect impacts.
AVERAGE ANNUAL ECONOMIC BURDEN, 2008-2010 ($)
Insulin pump 24,217.5
Insulin, no pump 30,103.2
Average savings 5,885.8
Technology-related impact per person affected* ($)
YEAR INSULIN PUMP INSULIN, NO PUMP ALL INSULIN
Treatment
expenditures
Indirect
impact**
Total
Treatment
expenditures
Indirect
impact**
Total
Treatment
expenditures
Indirect
impact**
Total
2005 2,544.3 19,716.1 22,260.5 3,451.9 24,068.9 27,520.8 3,408.1 29,699.2 33,107.3
2010 3,805.9 21,107.4 24,913.3 4,478.7 25,879.0 30,357.7 4,431.2 25,541.6 29,972.8
Average
(2008-2010)
3,863.8 20,353.7 24,217.5 4,471.5 25,631.7 30,103.2 4,431.0 25,283.9 29,714.9
* Person affected includes population reporting a condition, employed population reporting a condition, and employed caregivers by condition.
** A lower indirect impact value implies a greater contribution to the economy.
Insulin-related economic impact ($ millions)
Treatment
expenditures
Indirect
impact
Total
Treatment
expenditures
Indirect
impact
Total
Treatment
expenditures
Indirect
impact
Total
2005 424.2 1,093.4 1,517.6 11,338.6 26,310.6 37,649.2 11,762.7 27,404.1 39,166.8
2010 1,442.9 2,276.1 3,718.9 22,314.8 36,688.3 59,003.1 23,757.7 38,964.4 62,722.1
Average
(2008-2010)
1,223.1 1,993.2 3,216.4 20,002.4 35,425.3 55,427.7 21,225.6 37,418.6 58,644.1

Insulin dependent population affected (thousands)
PRC*
EPRC**
ECC***
PRC*
EPRC**
ECC***
PRC*
EPRC**
ECC***
2005 166.7 74.4 12.1 3,284.7 1,465.6 237.7 3,451.4 1,540.0 249.8
2010 379.1 146.5 25.6 4,982.4 1,925.6 336.4 5,361.5 2,072.2 362.0
* Population reporting a condition.
** Employed population reporting a condition.
*** Employed caregivers by condition.
Expenditures per PRC by site of service, 2010 ($)
OFFICE BASED OUTPATIENT INPATIENT EMERGENCY PRESCRIPTION HOME HEALTH TOTAL
Diabetes 608.8 851.8 21,435.3 1,073.9 1,212.6 5,636.3 2,330.4
Insulin 873.5 681.6 15,842.4 854.8 2,419.7 7,328.3 4,431.2
Insulin pump 436.7 340.8 6,020.1 179.5 3,278.2 3,664.2 3,805.9
Insulin, no pump 894.5 698.1 16,254.9 878.1 2,353.7 7,454.3 4,478.7
Diabetes population affected (thousands)
YEAR OVERALL DIABETES ALL INSULIN INSULIN-DEPENDENT DIABETES (%)
PRC*
EPRC**
ECC***
PRC*
EPRC**
ECC***
PRC*
EPRC**
ECC***
2005 17,019.9 7,219.0 1,171.0 3,451.4 1,540.0 249.8 20.3 21.3 21.3
2010 21,979.7 8,872.5 1,549.9 5,361.5 2,072.2 362.0 24.4 23.4 23.4
Economic impact associated with diabetes ($ millions)
YEAR DIABETES ALL INSULIN INSULINDEPENDENTDIABETES(%)
Treatment
expenditures
Indirect
impact
Total
Treatment
expenditures
Indirect
impact
Total
Treatment
expenditures
Indirect
impact
Total
2005 34,236.4 160,076.5 194,312.9 11,762.7 27,404.1 39,166.8 34.4 17.1 20.2
2010 51,222.5 208,750.2 259,972.7 23,757.7 38,964.4 62,722.1 46.4 18.7 24.1
Average
(2008-2010)
46,580.8 182,304.9 228,885.7 21,225.6 37,418.6 58,644.1 45.6 20.5 25.6

24 Healthy Savings
HEART DISEASE15
·· The average annual (2008-2010) economic burden associated
with using heart disease diagnostic tests and/or angioplasty
was $102.8 billion. Of this amount, direct treatment
expenditures added $62.6 billion to the health-care system,
and indirect impact accounted for $33.7 billion in lost GDP.
Further, the burden included an additional $6.5 billion related
to diagnostic tests performed on the healthy population.
(See Summary Chart: Heart Disease.)
·· For heart disease patients, the average annual (2008-2010)
savings per person affected was $1,930 compared to
those who did not use this technology. Although average
treatment expenditures were $4,534 higher for patients
using technology, the $1,930 savings stem from the $6,464
increase in GDP per person affected.
Heart disease is caused by the buildup of plaque in the arteries
near the heart, reducing blood flow. Potential consequences
include heart attack and heart failure. A range of technology
has been developed to mitigate the effects of heart disease,
from diagnostic tools such as EKG, echocardiograms, and chest
X-rays to therapeutic devices such as stents and pacemakers.
A substantial proportion of the heart disease population uses
these technologies, as seen in Figure 2.
15. Heart disease includes heart valve disorders, coronary atherosclerosis, cardiac dysrhythmias, myocardial infarction, and congestive heart
failure.
»» Heart disease is the leading cause
of death in the U.S.
• 600,000 deaths per year
• More than 700,000 Americans
suffer heart attacks annually.
• More than one in four heart
attack patients have had prior
heart attacks.
»» Risk factors include obesity,
aging, high blood pressure, high
cholesterol, and smoking.
Source: Centers for Disease Control and Prevention.
»» Angioplasty is a minimally invasive
procedure in which tubing is guided
through the coronary arteries with an
attached deflated balloon catheter.
Once the catheter reaches the blocked
artery, the balloon is inflated to widen
or unblock the artery. In some cases,
a stent is also inserted to reduce blockage.
»» Pacemakers may be used when the
heart beats too fast, too slow, or
irregularly. The small device, which is
implanted in the heart tissue, sends
electrical impulses that help the organ
beat regularly.

Figure 2
Proportion of heart disease patients using technology
Percent
34
36
38
40
42
44
46
2010
37.7
2009
39.0
2008
38.6
2007
40.6
2006
44.7
2005
43.8
We quantified the utilization of technology and economic effects for heart disease patients. Coronary events
related to this condition can hinder a patient’s ability to work, with 51 percent of heart attack patients returning
to their jobs within one month and 78 percent returning within six months.16
We used this information
to estimate lost workdays for heart disease patients in our calculation of indirect impact. Technology can
improve patients’quality of life and reduce presenteeism. According to Rosen et al.,17
surgical revascularization
represents a potential 22.4 percent quality of life increase if it prevents a major cardiac event. Presenteeism was
adjusted using this information.
PRC for heart disease in the United States expanded from 19.1 million in 2005 to 23.0 million in 2010. It is not
surprising that unhealthy lifestyles and demographic effects increased that population. In 2005, about
8.4 million (43.8 percent) people used technology, with an additional 250,000 users (37.7 percent) in 2010,
bringing the total to approximately 8.7 million. This decline in the percentage using technology may be
due to changes in insurance coverage or increased diagnosis of milder forms of the condition that require
management by medication only.
For this ailment, the aggregate economic burden increased from $220.1 billion in 2005 to $243.4 billion
in 2010. The burden associated with patients using technology was $87.8 billion in 2005, which rose to
$106.1 billion in 2010. The considerable increase in aggregate expenditures is due to PRC expansion for
heart disease overall.
The average annual (2008-2010) treatment expenditures per heart disease PRC from using technology
($7,050) were higher than for those who did not ($2,517), an indication of technology’s contribution to
rising health-care costs. However, many patients who underwent surgery survived solely as a result of that
costly method. Further, diagnostics can help in early detection and prevent expensive visits to hospitals
and emergency rooms. In fact, inpatient expenditures per PRC were lower for heart disease patients
using technology ($19,054) in 2010 compared to those who did not ($24,512). However, except for home
health-care services, all other sites of service were more expensive if they used technology. Some of these
differences in expenditures may be explained by the settings in which diagnostics were used. Such tests are
16. Amr E. Abbas, et al. “Frequency of Returning to Work One and Six Months Following Percutaneous Coronary Intervention for Acute Myocardial
Infarction,” American Journal of Cardiology 94, (2004).
17. Virginia M. Rosen et al. “Cost Effectiveness of Intensive Lipid-Lowering Treatment for Patients with Congestive Heart Failure and Coronary Heart
Disease in the U.S.,” Pharmacoeconomics 28, no. 1 (2010).

26 Healthy Savings
often undertaken during office visits and in emergency rooms. Diagnostic testing is part of the guidelines
for patients at risk for heart disease, and its absence might signal a lack of access to care and therefore
reduced spending. Additionally, patients using surgical technology may have more severe forms of heart
disease and may be more expensive to treat. This may contribute to the higher expenditures per PRC for
technology users.
Although heart disease technology could not contribute to savings to the health-care system between
2005 and 2010, productivity gains could offset some of the higher treatment costs, both during the
period of technology use and in the future. In 2010, the indirect impact for heart disease amounted to
$130 billion, but patients who used technology accounted for only $35 billion of that amount. For a better
understanding of these findings, the indirect impact per heart disease EPRC (or ECC, as appropriate) was
calculated. The average (2008-2010) indirect impact per person affected is much lower for technology
users than non-users. The improved labor market outcomes generated an additional $6,464 of GDP per
person affected in that period. For these individuals, screening may have allowed for better detection and
treatment of disease, and therapeutic technology may have reduced the time absent from work and raised
productivity as well.
One criticism that has been aimed at diagnostic technology is unnecessary application or overuse.
Health-care providers often recommend heart-related diagnostic tests for non-heart disease patients.
This type of expenditure added an average of $6.5 billion annually to the health-care system from 2008 to 2010.

SUMMARY CHART: HEART DISEASE
10
12
14
16
18
20
22
24
26
$ thousands
2005 2006 2007 2008 2009 2010
15.5
25.2
15.0
24.1
20.3
23.3
16.3
18.4
17.5
18.6 18.6
21.1
With technology Without technology
Economic impact of heart disease, 2005-2010*
Per person affected
AVERAGE ANNUAL ECONOMIC IMPACT, 2008-2010 ($)
With technology 17,441.3
Without technology 19,371.6
YEAR WITH TECHNOLOGY WITHOUT TECHNOLOGY TOTAL
Treatment
expenditures
Indirect
impact**
Total
Treatment
expenditures
Indirect
impact**
Total
Treatment
expenditures
Indirect
impact**
Total
2005 5,529.3 10,011.0 15,540.3 2,910.0 22,309.1 25,219.0 4,056.3 16,927.2 20,983.4
2010 7,407.5 11,163.1 18,570.6 2,958.6 18,118.7 21,077.4 4,635.0 15,497.8 20,132.8
Average
(2008-2010)
7,050.7 10,390.6 17,441.3 2,517.0 16,854.5 19,371.6 4,259.9 14,372.6 18,632.5

28 Healthy Savings
SUMMARY CHART: HEART DISEASE (continued)
Heart disease economic burden ($ millions)
Treatment
expenditures
Indirectimpact
Diagnosticsfor
healthypopulation
Total
Treatment
expenditures
Indirectimpact
Total
Treatment
expenditures
Indirectimpact
Diagnosticsfor
healthypopulation
Total
2005 46,411.6 35,185.4 6,231.4 87,828.4 31,389.0 100,921.7 132,310.7 77,800.6 136,107.1 6,231.4 220,139.1
2010 64,368.3 35,245.4 6,508.6 106,122.4 42,520.6 94,769.8 137,290.5 106,889.0 130,015.2 6,508.6 243,412.8
Average
(2008-2010)
62,604.4 33,684.9 6,522.5 102,811.8 35,844.7 89,055.1 124,899.8 98,449.1 122,740.0 6,522.5 227,711.6
Heart disease population affected (thousands)
PRC*
EPRC**
ECC***
PRC*
EPRC**
ECC***
PRC*
EPRC**
ECC***
2005 8,393.7 4,756.1 771.5 10,786.6 6,111.9 991.4 19,180.4 10,868.0 1,762.9
2010 8,689.6 4,329.4 756.3 14,371.7 7,160.4 1,250.8 23,061.3 11,489.8 2,007.1
Heart disease 792.6 2,721.0 20,831.1 1,838.8 565.0 5,736.1 4,635.0
Any technology 982.1 3,014.0 19,054.6 1,875.8 682.4 3,868.9 7,407.5
No technology 608.2 2,233.8 24,512.8 1,587.8 505.7 7,279.3 2,958.6

MUSCULOSKELETAL DISEASE
·· The average annual (2008-2010) economic burden
associated with using musculoskeletal disease-related
diagnostic tests and/or joint replacement surgery
expenditures added $23.1 billion to the health-care
system and $13.5 billion in lost GDP. Further, the
economic burden included an additional $8.3 billion in
diagnostic tests performed on the healthy population.
(See Summary Chart: Musculoskeletal Disease.)
·· For musculoskeletal disease patients, the average
annual (2008-2010) savings per person affected
were $24,518 compared to musculoskeletal disease
patients who did not use this technology. Although
average treatment expenditures were $3,887 higher
for musculoskeletal disease patients using technology,
the $24,518 savings arise from the additional $28,405
increase in GDP per person affected
Musculoskeletal disease is a chronic condition
that can disturb muscles, bones, and joints all over
the body and varies in severity. Musculoskeletal
diseases do not pose as high a mortality risk as
other prominent chronic illnesses, but they do
affect patients’ ability to perform the activities
of everyday living. To prevent the disease from
worsening, screening technologies for early
detection are often used. If it does worsen,
surgical procedures such as joint replacement
can greatly improve quality of life.
»» Musculoskeletal disease affected more
than 30 percent of the U.S. population
in 2006.
»» Arthritis, which constitutes a large
portion of musculoskeletal disease cases,
is a degenerative disease affecting nearly
30 percent of American adults in 2010.
»» Arthritis affects the non-elderly too.
In fact, two-thirds of people with
arthritis are under age 65.
»» Disability is high among rheumatoid
arthritis patients.
Sources: Centers for Disease Control and Prevention,
Burden of Musculoskeletal Diseases in the United States,
European Journal of Health Economics, Milken Institute.
»» MRI is a screening technique that can identify
bone erosions in arthritis earlier and with more
detail than typical X-rays.
»» Joint replacement surgeries involve removing
part or all of a damaged joint, such as a hip or
knee, and implanting a prosthesis.

30 Healthy Savings
Figure 3
Proportion of musculoskeletal disease patients using technology
Percent
9.0
9.5
10.0
10.5
11.0
11.5
12.0
10.9
11.5
11.2
10.0
11.0 10.9
201020092008200720062005
Technology can be very effective in improving outcomes for musculoskeletal disease patients. It can affect
health-care system costs as well as labor market outcomes. For joint replacement surgery, about 94 percent
of hip replacement patients return to work within two months, data shows, and the remaining 6 percent
return within a year.18
We used this data to calculate lost workdays. Presenteeism is also improved by surgery. David Ruiz and
colleagues estimated that knee replacement surgery added 3.4 quality-adjusted life years among patients
ages 40 to 44.19
Using this information, we adjusted presenteeism accordingly. Functional ability also
increases among joint replacement surgery patients, in the range of 56 to 79 percent.20
These positive effects
help to explain why technology has been consistently used by the musculoskeletal disease population.
With the aging of the population, rising obesity, and changing work environments, the musculoskeletal
disease PRC expanded from 26.3 million in 2005 to 41.1 million in 2010. Among them, about 2.9 million
used technology in 2005, which climbed to 4.5 million in 2010. While the number of patients treated with
technology increased, the percentage has remained relatively constant, around 10.9 percent.
Total treatment expenditures were $54.3 billion in 2005, climbing to $83.5 billion in 2010. Expenditures associated
with using technology were only $16.7 billion in 2005, and rose to $27.1 billion in 2010. These increases are
likely due to growth in the absolute PRC for both musculoskeletal disease in general and the technology
user population. The latter group comprises 10.9 percent of the PRC. The number varied through the six
years examined, but no trend is visible. Annual per-PRC treatment expenditures remained largely
unchanged during this period.
Average annual (2008-2010) expenditure per PRC was higher for technology users ($5,431) than non-users
($1,544), resulting in a loss to the health-care system of $3,887. Increased expenditures per PRC associated
with technology use were seen at all sites of service except home health care. Average home health
expenditures per person for patients using technology were $3,106, while the average for those without
technology was $5,180. Treatments are centered on supporting the activities of everyday living and may
18. Ryan M. Nunley et al. “Do Patients Return to Work After Hip Arthroplasty Surgery?” Journal of Arthroplasty 26, No. 6 Suppl. 1 (2011).
19. David Ruiz et al. “The Direct and Indirect Costs to Society of Treatment for End-Stage Knee Osteoarthritis,” Journal of Bone and Joint Surgery 95
(2013), pp. 1,473-1,480.
20. F. Cushner et al. “Complications and functional outcomes after total hip arthroplasty and total knee arthroplasty: Results from the Global
Orthopedic Registry (GLORY), The American Journal of Orthopedics 39, suppl. 9 (2010), pp. 22-28.

require close care by nurses; improved treatment would reduce this need. Less home care could also reflect
greater use of skilled nursing facilities after inpatient surgery, lowering per-PRC home-care expenditures.
In addition, surgery may be restricted to segments of the patient population. Those who already receive
frequent home care may be too frail to endure a surgical intervention, which may explain higher home
health expenditures per PRC in the non-technology user population.
In 2010, the indirect impact for musculoskeletal disease amounted to $608.9 billion. Because these ailments
affect the ability to perform daily activities, it follows that a large portion of the associated economic
burden would be generated by negative labor market outcomes. However, the indirect impacts for patients
treated with medical technology amounted to only $13 billion of that total. In 2010, the indirect impact was
$6,520 for those who used technology and $36,197 for those who did not, amounting to an additional gain
of $29,676 per person affected. For these individuals, treatment and screening may have shortened the time
absent from work due to musculoskeletal disease and improved productivity as well.

32 Healthy Savings
SUMMARY CHART: MUSCULOSKELETAL DISEASE
$ thousands
2005 2006 2007 2008 2009 2010
10
15
20
25
30
35
40
45
13.2
40.7
13.4
42.2
12.7
42.4
12.5
36.2
11.9
36.6
12.6
37.7
Withtechnology Withouttechnology
Economic impact of musculoskeletal disease, 2005-2010*
Per person affected
With technology 12,315.0
Without technology 36,832.9
Treatment
expenditures
Indirect
impact**
Total
Treatment
expenditures
Indirect
impact**
Total
Treatment
expenditures
Indirect
impact**
Total
2005 5,826.2 7,379.7 13,205.9 1,601.3 39,056.0 40,657.3 2,062.5 35,598.2 37,660.7
2010 6,049.0 6,520.7 12,569.7 1,541.2 36,196.6 37,737.7 2,032.8 32,959.9 34,992.7
Average
(2008-2010)
5,431.6 6,883.4 12,315.0 1,544.3 35,288.6 36,832.9 1,959.5 32,261.8 34,221.3

Musculoskeletal disease economic burden ($ millions)
Treatment
expenditures
Indirectimpact
Diagnostics
forhealthy
population
Total
Treatment
expenditures
Indirectimpact
Total
Treatment
expenditures
Indirectimpact
Diagnostics
forhealthy
population
Total
2005 16,733.6 13,435.2 6,226.8 36,395.6 37,532.7 585,649.7 623,182.4 54,266.4 599,084.9 6,226.8 659,578.0
2010 27,088.6 13,019.0 8,685.5 48,793.0 56,376.0 595,942.2 652,318.2 83,464.5 608,961.2 8,685.5 701,111.2
Average
(2008-2010)
23,103.4 13,473.4 8,301.6 44,878.4 54,913.4 587,038.9 641,952.3 78,016.8 600,512.3 8,301.6 686,830.7
Musculoskeletal disease population affected (thousands)
PRC*
EPRC**
ECC***
PRC*
EPRC**
ECC***
PRC*
EPRC**
ECC***
2005 2,872.1 2,461.2 399.2 23,438.7 20,085.5 3,258.1 26,310.8 22,546.7 3,657.4
2010 4,478.2 2,735.0 477.8 36,580.3 22,341.2 3,902.8 41,058.5 25,076.3 4,380.5
Musculoskeletal
disease
744.3 2,824.8 23,521.7 910.3 409.8 4,778.1 2,032.8
Any technology 1,401.4 3,062.5 25,126.6 978.9 322.4 3,106.3 6,049.0
No technology 650.9 2,683.2 22,168.9 897.5 421.9 5,180.0 1,541.2

34 Healthy Savings
COLORECTAL CANCER
·· The average annual (2008-2010) economic burden
associated with screening via colonoscopy/sigmoidoscopy
expenditures for colorectal cancer added $4.7 billion
to the health-care system and cost $12.6 billion in
lost GDP. Further, the economic burden included an
additional $17.4 billion in diagnostic tests performed
on healthy populations that revealed no polyps.
However, colonoscopy/sigmoidoscopy screening
prevented 560,000 people from developing the
illness, saving the health-care system $12.2 billion
and producing a gain to the economy. (See Summary
Chart: Colorectal Cancer.)
·· For colorectal cancer patients, the average annual
(2008-2010) savings per person affected were $97,302
compared to patients who had no screening.
In addition, screening helped in the prevention
of colorectal cancer, amounting to about $53,063
per case. In aggregate (including treatment and
prevention), the savings per person affected from
screening were $150,365. To assess the historical
trend of the effect of colorectal cancer screening on
the economic burden of the disease, it is necessary
to separate detection and prevention. For the
historical analysis, colorectal cancer patients who
had been screened were compared to those who
had not, according to current guidelines.21
We also assessed the number of colorectal cancer
cases prevented by screening and the expenditures
avoided. We determined the proportion of screening
performed on non-cancerous patients as well as the
proportion of polypectomies due to screening non-cancerous patients from the HCUP hospital database.
Assuming that one-third of growths removed in polypectomies would have turned into cancer, we applied
these proportions from HCUP to our findings on colorectal cancer patients screened from MEPS to
determine cases prevented and expenditures avoided.
21. The CDC recommends a colonoscopy every 10 years or a sigmoidoscopy every five years for people over 50.
»» Colorectal cancer is the third most
frequently diagnosed cancer in the
United States.
»» 80 percent of new cases occur in people
age 55 and over.
Health Economics.
»» Colonoscopy/sigmoidoscopy can detect and
remove polyps before they become cancerous.
»» Polyps can be removed by a polypectomy
procedure during colonoscopy. Although not
all polyps are cancerous, removing them can
prevent most colorectal cancer cases.
»» In 1988, only 27.8 percent of Americans age 50
and over had ever been screened, a proportion
that more than doubled to 65.7 percent by 2010.
»» 80 percent of reduced colorectal cancer incidence
is the result of increased screening.
Health Economics, Harvard University, Milken Institute.

In 2010, 617,000 Americans were treated for colorectal cancer, accounting for $3 billion in expenditures.
As with other diseases, MEPS was used to calculate the PRC for colorectal cancer patients who generated
screening and related expenditures. PRC with colonoscopies (adhering to national screening guidelines)
steadily increased from 459,300 in 2005 to 556,800 in 2010. Marketing campaigns aimed at increasing
prevention awareness among both providers and patients may have spurred adoption of this practice.
Overall expenditures per PRC were lower for those who had followed screening guidelines compared to
those who had not. This may be because screening catches cancer at an earlier stage, facilitating better
outcomes. In 2010, expenditures per PRC were $4,731 for patients with colorectal cancer, about $1,000 less
than those unscreened.
Productivity loss among colorectal cancer patients in the workforce is substantial, and labor market
participation is low. On average per annum (2008-2010), the indirect impact of colorectal cancer was
$22.9 billion. A person affected by the disease who had a screening added $96,399 to GDP annually
compared to the non-screened patient population. Detection at an early stage along with improved
treatments lead to better outcomes, which broadly lower absenteeism and presenteeism.
While the historical analysis included patients with colorectal cancer with and without colonoscopy,
the costs and effects of widespread colonoscopies on the healthy population must also be considered
as a consequence of increased technology adoption. In 2010, screening prevented 554,000 people from
developing colorectal cancer and saved $12 billion in health-care expenditures while increasing GDP.
However, an additional $17.7 billion was spent on screening the healthy population, who would not
obtain the disease.

36 Healthy Savings
SUMMARY CHART: COLORECTAL CANCER
$ thousands
2007 2008 2009 2010
19.3
153.2
20.3
148.5
16.5
189.2
17.3
167.4
With colonoscopy Without colonoscopy
0
20
40
60
80
100
120
140
160
180
200
Economic beneﬁt/loss associated with colonoscopy, 2007-2010*
Per person affected
With colonoscopy
(treatment and prevention)
18,030.1
Without colonoscopy 168,394.9
Economic impact per person affected* ($)
YEAR WITH COLONOSCOPY WITHOUT COLONOSCOPY TOTAL
Treatment
Prevention
Treatment
expenditures
Indirectimpact**
Total
Detection
Prevention
Treatment
expenditures
Indirectimpact**
Total
Treatment
expenditures
Indirectimpact**
Total
2005 11,871.1 64,022.8 75,893.9 - 12,280.6 100,968.8 113,249.4 11,911.4 72,858.2 84,769.6 -
2010 4,730.6 61,156.4 65,887.1 -48,584.8 5,893.2 161,554.6 167,447.7 4,843.1 85,165.9 90,009.0 -48,584.8
Average
(2008-
2010)
8,578.8 62,514.1 71,093.0 -53,062.9 9,482.3 158,912.7 168,394.9 8,723.6 85,567.1 94,290.8 -53,062.9

Total economic burden of prevention and treatment ($ millions)
YEAR WITH COLONOSCOPY WITHOUT COLONOSCOPY TOTAL
Treatment
Prevention
Treatment
expenditures
Indirectimpact
Total
Treatment
Prevention
Treatment
expenditures
Indirectimpact
Diagnostics
forhealthy
population
Total
Treatment
expenditures
Indirectimpact
Diagnostics
forhealthy
population
Total
2005 5,452.6 10,836.7 14,566.3 30,855.6 - 615.4 5,393.0 6,008.4 6,068.0 16,229.7 14,566.3 36,864.0 -
2010 2,634.2 15,324.0 17,659.3 35,617.5 -12,017.9 351.5 12,761.5 13,113.0 2,985.7 28,085.5 17,659.3 48,730.5 -12,017.9
Average
(2008-
2010)
4,711.2 12,557.2 17,445.1 34,713.4 -12,218.3 1,149.1 10,294.7 11,443.8 5,860.3 22,851.8 17,445.1 46,157.2 -12,218.3
Population affected by prevention and treatment (thousands)
TREATMENT
NUMBER
OF CASES
PREVENTED
WITHOUT COLONOSCOPY OVERALL TREATMENT
YEAR WITH COLONOSCOPY
PRC* EPRC** ECC*** PRC* EPRC** ECC*** PRC* EPRC** ECC***
2005 459.3 226.4 36.7 - 50.1 71.2 11.5 509.4 297.5 48.3
2010 556.8 339.5 59.3 554.4 59.6 106.7 18.6 616.5 446.2 77.9
OFFICE BASED OUTPATIENT INPATIENT PRESCRIPTION HOME HEALTH TOTAL
Overall colorectal cancer 1,124.9 7,068.1 20,877.8 532.5 8,324.0 4,843.1
With colonoscopy 934.5 7,198.5 24,281.6 605.5 8,324.0 4,730.6
Without colonoscopy 2,546.9 677.7 13,043.9 40.7 - 5,893.2

39
ECONOMIC IMPACT PROJECTIONS
AND MEDICAL TECHNOLOGY
M
edical device and technology advances exert economic impact in two primary ways: the expansion
effect and the substitution effect. Technology helps in detection and makes more patients suitable
for treatment, giving them a better chance of survival. As a result, the patient population increases,
creating the expansion effect. That leads to an increase in aggregate health-care costs, although the average
cost might fall due to fewer visits to expensive sites of service such as emergency rooms and hospitalization.
It also expands the workforce, resulting in economic growth. The substitution effect, on the other hand,
refers to newer technology supplanting older forms and influencing the unit cost of treatment.
As an illustration, let’s consider colorectal cancer screening by sigmoidoscopy/colonoscopy. With improved
technology for detecting polyps and malignant growths at earlier stages, along with greater efficacy and
safety, colorectal screening in the U.S. has grown tremendously in the past two decades. As screening
increased, incidence rates went up at first. However, the disease was detected earlier in many cases, likely
improving overall survival rates.
Percent
1988 1992 1996 2000 2004 2008
10
20
30
40
50
60
70
Figure 4
Colorectal cancer screening as proportion
of population 50+
Sources: Centers for Disease Control and Prevention, Milken Institute.
Per 100,000 population
1975 1980 1985 1990 1995 2000 2005 2010
40
45
50
55
60
65
70
Figure 5
Colorectal cancer incidence rates, age-
adjusted
Source: National Cancer Institute.
Incidence rates for colorectal cancer rose modestly from the mid-1970s through the mid-1980s as
colonoscopies identified more polyps and tumors, then those rates fell and have been declining ever
since. Coinciding with the drop in incidence rates, mortality has been declining since 1980, with the trend
accelerating since 1999.22
Thus, the initial rise in the incidence of colorectal cancer, or expansion effect,
is attributed to early detection and increased survival.
22. National Cancer Institute.

40 Healthy Savings
Moreover, improved technology is regularly substituted for older methods in treating established patients.
The unit cost of new technologies may be higher or lower than those they replace. However, along with
fostering health improvements, technology can curtail visits to expensive sites of service, such as hospitals
and emergency rooms. In the case of heart attacks, the chances of survival depend on the successful
opening of blocked arteries. In the late 1960s, bypass surgery, a major open-heart procedure, saved lives
by grafting an artery or vein around the occluded coronary artery. An improved technology known as
angioplasty was developed in the late 1970s, involving the use of a balloon catheter to break up the
blockage.
Since the mid-1990s, angioplasty has increasingly incorporated the insertion of stents—small mesh
tubes that hold the coronary artery open—in the area of the blockage. Later generations of stents have
reduced mortality and improved overall outcomes. Further, as the technology has become less invasive,
quality adjusted life year (QALY) has improved for heart disease patients. In fact, a study points out
that approximately 70 percent of survival improvement is the result of progress in technology, with the
remainder stemming from changes in risk factors such as smoking.23
With technology expanding the patient population due to early detection and better survival outcomes,
it also increases aggregate treatment expenditures, even assuming constant per-patient expenses. On the
other hand, increased survival means more people in the workforce. Less invasive technology may ease
average treatment expenditures (depending on the disease), perhaps offsetting some of the increases
discussed above. Further, this factor helps worker productivity, supporting the labor market.
Knowing how medical innovations affect future treatment expenditures is fundamental to making prudent
investment decisions in the field. It’s also essential to understand how a particular technology contributes to
or detracts from the GDP. One objective of this report is to project the overall economic impact associated
with medical technologies through 2035. Further, our report provides data-driven evidence for stakeholders
to discern the ties between innovations and disease-specific economics. With this in mind, we simulate
three future innovation scenarios—which influence the utilization and diffusion of medical technology—
and project the economic impact associated with each.
1) Continued incentives (baseline): In this scenario, the growth in medical innovation remains at the same
historical pace, along with the growth rate of its use.
2) Increased incentives (optimistic): Medical innovation advances at a higher than historical rate.
3) Decreased incentives (pessimistic): Medical innovation progresses at a lower than historical rate.
23. David M. Cutler. “The Lifetime Costs and Benefits of Medical Technology,” NBER Working Paper Series (2007).

41Economic Impact Projections and Medical Technology
Projection of PRC
To estimate future treatment expenditures and indirect impact, we first projected the PRC and integrated
other relevant data. An appropriate model for the projection of PRC associated with disease-specific
technology involves a range of decision-making stages and options. We used decision trees that illustrate
health processes over time to create disease-specific Markov models.
To elaborate, let’s study the effect of disease A on a hypothetical cohort of 100 people in 2010 using a
Markov model. Any individual can be well or have disease A. If they suffer from disease A, they can have
either the mild or severe form. Suppose 50 are well, 25 have a mild form of the disease, and 25 have the
severe form. Every year, of course, people in the cohort can remain in the same health state, transition into
a different one, or die, and the likelihood of each event can be estimated.
Figure 6
A basic Markov decision tree
Well
Survive
Die, other causes
Survive
Die, other causes
or mild disease
Dead
Severe disease
Dead
Dead
Mild disease
Severe disease
Dead
Maintain well
Develop disease
Progress to
severe disease
Maintain mild
disease
Well
Mild disease
Mild disease
Severe disease
Survive
Die, other causes
or severe disease
M
That includes the 50 individuals who are well. Assuming a 2 percent probability of death, one of them will die
by the next year. Assuming a 10 percent chance of getting mild disease, about five of the remaining 49 people
would acquire the ailment. The rest of the 44 people will remain well for the next year. Other branches of the
decision tree such as“mild disease”or“severe disease”will follow a similar logic.
This simplified model serves as a basis for even the most complex Markov model. Ours are based on the
biological progression and treatment patterns for the diseases we examined. The probabilities of the events
in our models were derived from the scientific literature and MEPS data.

42 Healthy Savings
Projection of treatment and prevention expenditures
The projected PRC was multiplied by expenditures per PRC (from MEPS) for each health state to estimate
the projected expenditure for each health state. Expenditures for each health state were aggregated to
calculate total projected expenditures for each disease. We use MEPS expenditures per PRC because they
incorporate costs for six sites of service related to the assessed diseases. However, we could not incorporate
costs for associated diseases, such as diabetes’ role as a risk factor for heart disease.
For the increased incentives scenario, the annual reduction in expenditures per PRC is applied to diabetes,
heart disease, and musculoskeletal disease to account for reductions in complications and use of expensive
sites of service associated with better technology use. Similarly, an annual percentage increase in expenditures
per PRC is applied to the decreased incentives scenario. An annual percentage change is not assumed for
colorectal cancer as it is for the other diseases because the duration and intensity of treatment vary widely
depending on the stage of the disease.
With risk factors such as aging and obesity projected to increase with time, the PRC is expected to suffer more
severe disease. Without improvements in medical technology, patients will make more ER visits, have more
frequent complications, and be more expensive to treat on average. The difference among economic impact
scenarios demonstrates benefits and losses associated with investing in technology innovation.
The following table shows the projected treatment expenditures associated with innovation through 2035,
in 2010 dollars using a discount rate of 3 percent.
Table 6
Projected treatment expenditures by disease
2010-2035 ($ billions*)
ABSOLUTE DIFFERENCE
Continued
incentives
Increased
incentives
Decreased
incentives
Continued-
increased
Continued-
decreased
Diabetes 1,622.4 1,602.8 1,631.7 19.6 -9.3
Heart disease 2,663.6 2,628.2 2,853.1 35.4 -189.5
Musculoskeletal disease 1,983.0 1,952.4 2,014.6 30.6 -31.5
Colorectal cancer 214.7 204.1 220.7 10.6 -6.0
Colorectal cancer prevented -120.5 -137.2 -109.2 16.7 -11.2
* In 2010 dollars.
Sources: Medical Expenditure Panel Survey, Milken Institute.

In 2010 dollars, improved technology (following the increased incentives scenario compared to continued
incentives) for insulin pumps can save the health-care system $19.6 billion. However, lower incentives in
device technology would increase costs by $9.3 billion. For heart and musculoskeletal diseases, the gain to
the health-care system is $35.4 billion and $30.6 billion, respectively. Decreased innovation would raise care
expenditures by $189.5 billion for heart disease and $31.5 billion for musculoskeletal disease. For colorectal
cancer, the savings associated with innovation stem from early detection and prevention. Better technology
can detect polyps earlier and remove them, preventing cancer. Aggregate savings to the health-care system
due to better diagnostics are $27.3 billion, whereas lowered incentives to screen will increase the incidence
of cancer, adding $17.2 billion in expenditures.
Projection of indirect impact (foregone GDP)
Improved technology also has profound labor market implications. Due to early detection, prevention,
and higher quality of life, work outcomes often greatly improve for people affected by these diseases.
For a comprehensive analysis, we also computed the gain and loss to GDP associated with each technology
incentives scenario.
We estimated the population reporting a condition for each disease through 2035. Further, projected PRC
and U.S. employment data were used to calculate employed population reporting a condition projections.
Projections of employed caregivers by condition are proportional to the EPRC estimations. Similar to
the historical trend methodology, a GDP-based approach was used to estimate relevant indirect impact.
Except for colorectal cancer, the indirect impact in the increased incentives scenario was adjusted further
downward (assumptions are similar to those used in the historical methodology) due to improved labor
market outcomes. The indirect impact for decreased incentives was adjusted upward due to a projected
increase in the severity of chronic disease and negative labor market effects. We did not adjust colorectal
cancer’s indirect impact due to variation in the severity and length of the disease.
The cumulative GDP gain in 2010 dollars associated with accelerated technology innovations in the
increased incentives scenario (compared to continued incentives) is $205.8 billion for diabetes,
$773.7 billion for heart disease, and $250.4 billion for musculoskeletal diseases. The contribution to
GDP from colorectal cancer patients was $109 billion, and because screening also spares many people
from cancer, $41.8 billion more was added to the economy.
However, considering the decreased incentives scenario (compared to continued incentives), diabetes
reduced GDP by $91.8 billion. Similarly, decreased incentives lead to a GDP loss of $1.4 trillion for heart
disease and $277.2 billion for musculoskeletal disease. Treatment and prevention of colorectal cancer
reduced GDP by $94.5 billion.

44 Healthy Savings
Table 7
Projected foregone GDP by disease
2010-2035 ($ billions*)
ABSOLUTE DIFFERENCE
Continued
incentives
Increased
incentives
Decreased
incentives
Continued-
increased
Continued-
decreased
Diabetes 10,720.2 10,514.4 10,812.0 205.8 -91.8
Heartdisease 5,073.7 4,300.0 6,435.5 773.7 -1,361.8
Musculoskeletaldisease 22,690.5 22,440.0 22,967.7 250.4 -277.2
Colorectalcancer 1,790.5 1,681.5 1,851.9 109.0 -61.5
Colorectalcancerprevented -331.5 -373.3 -298.5 41.8 -33.0
* In 2010 dollars.
The following subsections elaborate on the methods and other findings for each disease. For further
information, please see the Methodology section of this report.
DIABETES
We created a Markov model that assumes a “well” initial state for a cohort of individuals (which includes all
undiagnosed diabetics in the U.S.) and follows them over 25 years. Many are later diagnosed with diabetes
and progress through the disease. As noted earlier, there are two types of diabetes. Type 1, often referred to
as juvenile diabetes, is an autoimmune disease with an earlier onset. Type 2 is more common, with obesity,
aging, and high cholesterol as risk factors. Type 1 diabetes generally requires insulin use upon diagnosis,
and type 2 diabetes requires insulin treatment after reaching a certain level of severity. For the purposes of
this model, the two types of diabetes were combined, distinguishing instead between insulin dependent
and non-insulin dependent diabetes.
With the progression of the disease, many non-insulin dependent diabetics transition into insulin dependence.
Injections and pumps are the most common modes of administering insulin.Those using injections can begin
to use pumps at a certain point in this framework; however, once pump use is initiated, it was assumed to
continue throughout the patient’s life.
While anyone in the model is subject to mortality risk, diabetes poses an increased risk of death. Among
diabetics, non-insulin dependent patients have a less severe form and are less subject to complications and
hypoglycemic events. Therefore, their risk of death is lower than insulin-dependent diabetics. Pumps reduce
the likelihood of such events by maintaining a“healthy”blood sugar level, resulting in a lower mortality risk
for users than for patients who inject insulin.
The main difference among the scenarios is the probabilities associated with the initiation of pump use and
the consequences of better disease management. It is assumed that improved technology will expand the
population suitable for using pumps and that a higher proportion will adopt the technology. The opposite is

true in the decreased incentives scenario. As such, the PRC for increased incentives assumed a higher annual
takeup rate for insulin pumps, twice that of continued incentives, and the PRC for decreased incentives
assumed a lower rate of use.
Further, an annual percentage reduction was applied to the per-PRC expenditures and the indirect impact due
to improved disease management in the increased incentives scenario. Similarly, a percentage increase in those
expenditures and indirect impact was applied to the decreased incentives scenario for the opposite reason.
Analysis of the model over a 25-year period reveals that the population reporting a condition for diabetes
was 22 million in 2010, which is projected to rise to 55.6 million by 2035 in the continued incentives
scenario. This dramatic increase can be attributed to the rising average age of Americans and the widening
prevalence of obesity and high cholesterol. The overall diabetes PRC increases slightly in the increased
incentives scenario due to the lower mortality risk associated with increasing pump use.
The increased incentives scenario is projected to have approximately 50,000 more diabetic PRC in 2035
compared to continued incentives, while decreased incentives projects 30,000 fewer. The primary change
in PRC comes from changes in the population of pump users among the scenarios, representing a relatively
small proportion of the overall PRC. The larger diabetic PRC in the increased incentives scenario arises from
a reduction in deaths by virtue of improved care. Under decreased incentives, the narrower PRC stems from
the larger number of diabetes deaths.
Compared to continued incentives, the increased incentives scenario expands insulin pump use robustly by
2035. Similarly, the decreased incentives scenario has about half the number of pump users associated with
continued incentives.This follows the assumptions about technology adoption in each projection.The PRC of
non-insulin dependent diabetics also remains constant across scenarios and accounts for the largest portion
of diabetics. The insulin dependent category has a lower PRC because it is typically associated with the rarer
auto-immune-related type 1 disease and more severe type 2.
Diabetes direct treatment expenditures were $51 billion in 2010.The continued incentives scenario increases
that to $131.4 billion in 2035. (See Projections: Diabetes.) Expenditures over 25 years are $19.6 billion less
for the increased incentives scenario and $9.3 billion more for the decreased incentives scenario compared
to the sum for continued incentives in 2010 dollars. Although overall diabetes PRC rises in the increased
incentives scenario, total expenditures decrease due to the lower average expenditures per PRC associated
with insulin pump use. Similarly, overall expenditures grow in the decreased incentives scenario due to larger
expenditures per PRC.
Under the continued incentives assumption that medical technology applied to diabetes will steadily
advance, the total indirect impact will leap from $208.8 billion in 2010 to $1.2 trillion in 2035. With the
technology assumptions in the increased incentives scenario, indirect impact will also continue to increase
through 2035. However, compared to the continued incentives scenario, it will produce cumulative savings
of $354.4 billion. In 2010 dollars, that amounts to $205.8 billion.
These savings may be a result of better disease management as well as technology adoption, both of which
can make for a healthier and more productive workforce. Compared to the continued incentives scenario,
decreasing incentives for medical technology contributes to $162.8 billion in productivity loss, or $91.8 billion
in today’s dollars. Worse labor market outcomes may be attributed to more severe disease in the employed
population and a relative lack of treatment options due to weaker innovation.

46 Healthy Savings
PROJECTIONS: DIABETES
Increased Incentives
compared to continued incentives
Decreased Incentives
Projected savings†
: Diabetes
0
5
10
15
20
25
30
35
40
45
$ billions
0.00
2.74
7.85
15.76
26.97
42.13
2010 2015 2020 2025 2030 2035
-25.0
-20.0
-15.0
-10.0
-5.0
0.0
$ billions
-23.52
-12.65
-6.35
-2.79
-0.84
0.00
2010 2015 2020 2025 2030 2035
Economic impact of diabetes, 2010-2035 ($ billions)†
CUMULATIVE IMPACT INCREASED INCENTIVES DECREASED INCENTIVES
Direct expenditures 32.5 -15.3
Gain/loss to the economy 354.4 -162.8
Due to survival 0.354 0.081
Additional gain/loss 354.1 -162.8
Total 387.0 -178.1
Projected diabetes population affected (millions)
PRC*
ABSOLUTE
DIFFERENCE
EPRC**
ABSOLUTE
DIFFERENCE
ECC***
ABSOLUTE
DIFFERENCE
Year
Continuedincentives
Increasedincentives
Decreasedincentives
Continued-increased
Continued-decreased
Continuedincentives
Increasedincentives
Decreasedincentives
Continued-increased
Continued-decreased
Continuedincentives
Increasedincentives
Decreasedincentives
Continued-increased
Continued-decreased
2010 21.98 21.98 21.98 0.00 0.00 8.87 8.87 8.87 0.00 0.00 1.55 1.55 1.55 0.000 0.000
2035 55.59 55.65 55.57 -0.05 0.03 22.66 22.68 22.65 -0.02 0.01 3.958 3.962 3.956 -0.004 0.002

Projected economic impact of diabetes ($ billions) †
TREATMENT EXPENDITURES
ABSOLUTE
DIFFERENCE
INDIRECT IMPACT
ABSOLUTE
DIFFERENCEYear
Continued
incentives
Increased
incentives
Decreased
incentives
Continued-
increased
Continued-
decreased
Continued
incentives
Increased
incentives
Decreased
incentives
Continued-
increased
Continued-
decreased
2010 51.0 51.0 51.0 0.0 0.0 208.8 208.8 208.8 0.0 0.0
2035 131.4 128.6 132.7 2.9 -1.3 1,212.9 1,173.6 1,235.1 39.3 -22.2
Cumulative
(2010-2035)
2,430.8 2,398.3 2,446.2 32.5 -15.3 16,832.0 16,477.6 16,994.8 354.4 -162.8
In2010dollars
(2010-2035)
1,622.4 1,602.8 1,631.7 19.6 -9.3 10,720.2 10,514.4 10,812.0 205.8 -91.8
† Screening expenditures for the healthy population not included.
HEART DISEASE
Heart disease involves narrowing of the blood vessels around the heart, reducing blood flow and
oxygen supply. The consequences can be serious, including heart attack or cardiac arrest. The disease
process was modeled in a Markov model first, followed by the effects of technology. Echocardiogram,
electrocardiogram (EKG), and chest X-ray were analyzed as diagnostic tools, and angioplasty and
pacemaker insertion as surgical treatments.
The incidence of heart disease is affected by a variety of risk factors, including age, smoking status,
diabetes, high cholesterol, obesity, and gender. For the purpose of this model, the risks included were aging
and obesity,24
two of the most significant conditions affecting the development of the disease. With these
factors projected to increase over time in the United States, we incorporated that likelihood into the model.
Incidence was calculated from Framingham risk prediction models for coronary heart disease using data from
the Centers for Disease Control and Prevention and the National Health and Nutrition Examination Survey.
The influence of other risks or variations in the trajectory of incidence is assessed in sensitivity analysis.
Heart disease can present with symptoms, primarily angina pectoris or chest pain, but oftentimes it is
present without.Therefore there is an undiagnosed heart disease health state within the model. If disease is
identified, depending on the probability of screening or identification of symptoms, a clinician can prescribe
medication and lifestyle changes that can slow or stop its progression. Undiagnosed, untreated heart
disease can pose high risk for acute side effects. Improvements in diagnostic testing technologies such as
EKG, echocardiogram, or chest X-ray may improve a clinician’s ability to identify and subsequently treat the
condition, so we changed the sensitivity of diagnostic testing among incentive scenarios.
24. Smoking was not included because smoking initiation has been decreasing in the United States and might not play a significant role in
projected incidence.

48 Healthy Savings
As the disease increases in severity, it raises the probability of acute coronary events such as myocardial
infarction (heart attack) and cardiac arrest, both of which can be fatal. It was assumed that heart disease
would be diagnosed by symptom identification after such an occurrence. Acute coronary events are
expensive, often requiring emergency room and inpatient care, and potentially surgery. Surgery can also
be planned (if heart disease is diagnosed) to prevent such an occurrence. While surgery can curb the future
consequences of the disease and reduce chance of restenosis (blockage of the artery), the procedure itself
involves the risk of death. However, with improvements in technology, the risk of death declines, which is
also incorporated into the model.
The main differences among scenarios
are the likelihood of undergoing planned
surgery and diagnostics, the risk of death
from surgery, overall heart disease death
rates, the likelihood of early detection using
diagnostics, and improved treatment. In the
increased incentives scenario, there is more
innovation and diffusion of technology
through the medical field.Therefore, higher
rates of adoption of both diagnostic and
therapeutic technology was assumed.
Similarly, lower rates of technology use were
assumed in the decreased incentives scenario.
Because the technology was assumed to improve with more innovation in the increased incentives
scenario, the accuracy of technology and its ability to inform proper treatment methods were assumed to
improve. With better technology, the risk of death tied to surgery was assumed to decrease. Without such
improvements, the risk of death would not decrease.
Changes in expenditures per PRC and indirect impact per person affected were also changed across the
projections. They were adjusted downward in the increased incentives scenario to account for reduced
complications and increased productivity. Similarly, expenditures per PRC and indirect impact per person
affected rose to address a lack of adequate treatment options for more severe cases.
The population reporting a condition for heart disease was calculated at 23.1 million in 2010, projected to
increase approximately 68 percent to 38.9 million in 2035. (See Projections: Heart Disease.) The increased
incentives scenario reveals a rise in PRC over time, totaling 41.8 million in 2035, while decreased incentives
reveals a PRC of 38.3 million. Increased adoption of testing technology allows more people to be diagnosed
with the disease. Combined with surgeries that potentially prevent fatal coronary events, an increased
PRC reveals greater access and higher quality of care. Fewer PRCs associated with the decreased incentives
scenario corresponds to an increase in undiagnosed disease and incidence of death.
Heart disease treatment expenditures totaled $106.9 billion in 2010 and will increase to $180.2 billion in
2035 for the continued incentives scenario. Increased incentives will reduce aggregate expenditures
$81.4 billion more than the continued incentives scenario over 25 years, equivalent to $35.4 billion in 2010
dollars. Initially the increased incentives scenario is more expensive due to a larger population of diagnosed
patients obtaining treatment and reduction in mortality. However, the expenditures per PRC shrink,
contributing to a cumulative savings in the 25-year period. The rising expenditures associated with the
Over 25 years, the decreased
incentives scenario results in
a $316.4 billion expansion in
treatment expenditures for
heart disease, or $189.5 billion
in 2010 dollars.

increased incentives scenario correspond to an increase in proper treatment and longer lives, both positive
outcomes not directly measured in this study.
The decreased incentives scenario sees higher costs than the continued incentives scenario because,
while fewer patients are obtaining treatment and fewer are alive, we assume an increase in per-PRC
expenditures. Over 25 years, the decreased incentives scenario results in a $316.4 billion expansion in
treatment expenditures, or $189.5 billion in 2010 dollars.
Because some of the technologies assessed include diagnostic tests, the costs of screening the healthy
population must also be considered. In the continued incentives scenario, such screening costs the health-
care system $238.5 billion cumulatively over 25 years. Higher screening rates in the increased incentives
scenario result in an additional $21.2 billion expenditure, while the decreased incentives scenario saves
$5.2 billion. This additional expenditure was not included in the economic impact estimates because these
people do not have the examined diseases.
Tied to the changes in PRC, the indirect impact for heart disease will be substantial over the 25-year period
and is expected to more than double under all scenarios. With the introduction of new technology under
increased incentives, indirect impact will be lower amid an expansion of the labor force (due to survival) and
productivity tied to improved quality of life. This will generate a cumulative $1.3 trillion gain in GDP over
25 years, $44.2 billion of which can be attributed to improved survival. This is equivalent to a $773.7 billion
gain in 2010 dollars.
Decreased incentives will result in a GDP loss of $2.4 trillion, or $1.4 trillion in 2010 dollars, due to drained
productivity and a decreased EPRC as more people die or exit the labor market.

50 Healthy Savings
PROJECTIONS: HEART DISEASE
: Heart disease
-400
-350
-300
-250
-200
-150
-100
-50
0
0.00
-13.48
-44.39
-98.56
-192.73
-354.69
$ billions
2010 2015 2020 2025 2030 2035
-10
10
30
50
70
90
110
0.00
14.36
40.68
63.09
86.25
113.33
2010 2015 2020 2025 2030 2035
$ billions
Economic impact of heart disease, 2010-2035 ($ billions)†
Gain/loss to the economy 1,263.0 -2,409.9
Additional gain/loss 1,218.8 -2,487.0
Total 1,344.4 -2,726.4
Projected heart disease population affected (millions)
PRC*
ABSOLUTE
DIFFERENCE
EPRC**
ABSOLUTE
DIFFERENCE
ECC***
ABSOLUTE
DIFFERENCE
Year
Continuedincentives
Increasedincentives
Decreasedincentives
Continued-increased
Continued-decreased
Continuedincentives
Increasedincentives
Decreasedincentives
Continued-increased
Continued-decreased
Continuedincentives
Increasedincentives
Decreasedincentives
Continued-increased
Continued-decreased
2010 23.1 23.1 23.1 0.0 0.0 11.5 11.5 11.5 0.0 0.0 2.0 2.0 2.0 0.0 0.0
2035 38.9 41.8 38.3 -2.9 0.6 19.6 21.0 19.3 -1.5 0.3 3.4 3.7 3.4 -0.3 0.1

Projected economic impact of heart disease ($ billions)†
ABSOLUTE
DIFFERENCE
INDIRECT IMPACT
ABSOLUTE
DIFFERENCEYear
Continued
incentives
Increased
incentives
Decreased
incentives
Continued-
increased
Continued-
decreased
Continued
incentives
Increased
incentives
Decreased
incentives
Continued-
increased
Continued-
decreased
2010 106.9 106.9 106.9 0.0 0.0 130.0 130.0 130.0 0.0 0.0
2035 180.2 166.5 209.1 13.7 -28.9 508.0 408.4 833.7 99.6 -325.7
Cumulative
(2010-2035)
3,876.0 3,794.6 4,192.5 81.4 -316.4 7,781.5 6,518.5 10,191.5 1,263.0 -2,409.9
In 2010 dollars
(2010-2035)
2,663.6 2,628.2 2,853.1 35.4 -189.5 5,073.7 4,300.0 6,435.5 773.7 -1,361.8
Projected expenditures on healthy population screening/diagnostics
HEALTHY PEOPLE SCREENED
(MILLIONS)
SCREENING EXPENDITURES
($ BILLIONS)
Year
Continued
incentives
Increased
incentives
Decreased
incentives
Continued
incentives
Increased
incentives
Decreased
incentives
2010 21.3 21.3 21.3 6.5 6.5 6.5
2035 22.7 26.4 21.8 11.9 13.9 11.4
Cumulative
(2010-2035)
587.1 634.2 575.2 238.5 259.7 233.3
†Screening expenditures for the healthy population not included.
MUSCULOSKELETAL DISEASE
Musculoskeletal disease encompasses a range of conditions. In general, it is a chronic, progressive disorder of
the joints that affects quality of life and is associated with a small increase in risk of death. The musculoskeletal
disease Markov model was created to assess the economic effects of device innovation in medical technology.
Specifically evaluated were MRI for diagnosis and joint replacement surgery as a treatment.
Rheumatoid and osteoarthritis were used as the primary proxies for the category during the modeling
process. They are assumed to begin as mild disease and progress to more severe, debilitating disease that
may require more drastic surgical treatment or may render a patient disabled.
The average of the incidence rates for rheumatoid and osteoarthritis was matched with historic MEPS
data and used as the incidence for musculoskeletal disease. Since musculoskeletal disease becomes
more common with age and joints become more strained with higher body weight, aging and obesity

52 Healthy Savings
were used as risk factors for increasing incidence. MRI was assessed as a potential diagnostic tool and
proper identification of the disease was assumed to lead to treatment, be it a medical intervention or a
lifestyle modification to reduce joint stress.
Once the disease progresses to a severe stage, a patient might require surgery. Such a procedure can
succeed in relieving symptoms or may require a revision. In case of an unsuccessful revision surgery,
treatment failure is assumed.
The main differences among the incentive scenarios are the likelihood of obtaining a diagnostic test and
its efficacy in identifying disease, the likelihood of having surgery, the relative risk of progression from mild
to severe disease with treatment, the revision rate (likelihood of requiring additional surgery) and surgical
mortality rate, and the relative risk of death due to disease. We use historical trends from MEPS to inform the
likelihood of diagnostic testing and surgery, and reviewed the literature to assign value to other variables.
In the increased incentives scenario, rising innovation is assumed to result in better, more accurate diagnostic
and surgical technology. We project a higher rate of increase of technology use over time compared to the
continued incentives scenario. As diagnostic accuracy rises, treatment improves due to better diagnostics
and the mortality and revision rates decline. With improved technology reducing long-term complications
and improving the ability to perform everyday living activities, the death rate due to musculoskeletal disease
eases slightly in the increased incentives scenario. Under decreased incentives, technology is assumed to
develop more slowly. The accuracy of diagnostic tools and the surgical mortality and revision rates all remain
the same. We project a lower rate of increase for technology use compared to the continued incentives
scenario. Because technology improves slowly as disease severity worsens, the effectiveness of current
treatments declines and the risk of death from musculoskeletal disease slightly rises.
Musculoskeletal disease had a PRC of 41.1 million in 2010, which increases to 66.6 million by 2035 in the
continued incentives scenario, primarily due to aging and obesity. (See Projections: Musculoskeletal Disease.)
Increased incentives yields a slightly greater PRC of 67.3 million. That scenario restrains the disease from
progressing in severity, which is associated with slightly higher mortality. A reduction in severity lowers
death rates but increases overall PRC.
The PRC for musculoskeletal disease does not appear to vary widely among the incentive scenarios because
the illness does not greatly increase risk of death. It does affect quality of life, however, and if PRC were
adjusted to account for that, the differences would be more apparent. Additionally, the increased incentives
scenario halves the number of people with undiagnosed disease or deprived of proper treatment by 2035.
This represents a substantial improvement in care.
Total expenditures for musculoskeletal disease were $83.5 billion in 2010, increasing to $135.4 billion in
2035 in the continued incentives scenario. Expenditures for increased incentives are $131.3 billion by
2035, saving a cumulative $50.3 billion compared to continued incentives, or $30.6 billion in 2010 dollars.
Because the PRC is higher compared to the increased incentives scenario, the savings mainly arise from the
reduction in annual expenditures per PRC associated with improved technology. The decreased incentives
projection produces a $52 billion increase in cumulative expenditures over 25 years due to the higher
treatment costs associated with more severe disease. This equals $31.5 billion in 2010 dollars.
The more frequent use of MRI as a diagnostic technique will increase the likelihood that a healthy person
is screened, incurring the cost of use. In 2010, $8.7 billion was spent on diagnostic MRI for the healthy
population. Using rates of diagnostic testing from the model, the continued incentives scenario results in

a cumulative burden of $375.6 billion over 25 years. Increasing diagnostic use in the increased incentives
scenario results in $473.6 billion, while decreased incentives creates a cumulative burden of $326.6 billion
for screening the healthy population.
Indirect impact for the EPRC and caregivers was also examined.The savings to GDP involving musculoskeletal
disease will be substantial over the 25-year period. While the indirect impact will grow under all scenarios,
progress in technology and improved survival under the increased incentives scenario will expand the
workforce and reduce indirect impact. These advances will improve quality of life, which will also raise labor
market participation and productivity. Cumulatively, GDP will benefit by $393.9 billion, or $250.4 billion in
2010 dollars.

54 Healthy Savings
PROJECTIONS: MUSCULOSKELETAL DISEASE
: Musculoskeletal disease
2010 2015 2020 2025 2030 2035 2010 2015 2020 2025 2030 2035
$ billions$ billions
0
5
10
15
20
25
30
-80
-70
-60
-50
-40
-30
-20
-10
0
0.00
0.00
-3.33
-9.34
-19.68
-37.51
-67.99
8.98
16.18
21.49
25.17
27.16
Economic impact of musculoskeletal disease, 2010-2035 ($ billions)†
Total 444.2 -538.5
Projected musculoskeletal disease population affected (millions)
PRC*
ABSOLUTE
DIFFERENCE
EPRC**
ABSOLUTE
DIFFERENCE
ECC***
ABSOLUTE
DIFFERENCE
Year
Continued
incentives
Increased
incentives
Decreased
incentives
Continued-
increased
Continued-
decreased
Continued
incentives
Increased
incentives
Decreased
incentives
Continued-
increased
Continued-
decreased
Continued
incentives
Increased
incentives
Decreased
incentives
Continued-
increased
Continued-
decreased
2010 41.1 41.1 41.1 0.0 0.0 25.1 25.1 25.1 0.0 0.0 4.4 4.4 4.4 0.0 0.0
2035 66.6 67.3 65.7 -0.7 0.9 41.1 41.6 40.5 -0.4 0.6 7.2 7.3 7.1 -0.1 0.1

Projected economic impact of musculoskeletal disease ($ billions)†
ABSOLUTE
DIFFERENCE
INDIRECT IMPACT
ABSOLUTE
DIFFERENCEYear
Continued
incentives
Increased
incentives
Decreased
incentives
Continued-
increased
Continued-
decreased
Continued
incentives
Increased
incentives
Decreased
incentives
Continued-
increased
Continued-
decreased
2010 83.5 83.5 83.5 0.0 0.0 609.0 609.0 609.0 0.0 0.0
2035 135.4 131.3 139.8 4.1 -4.4 2,289.4 2,266.3 2,353.0 23.1 -63.6
Cumulative
(2010-2035)
2,889.7 2,839.4 2,941.7 50.3 -52.0 34,854.8 34,460.9 35,341.2 393.9 -486.4
In 2010 dollars
(2010-2035)
1,983.0 1,952.4 2,014.6 30.6 -31.5 22,690.5 22,440.0 22,967.7 250.4 -277.2
HEALTHY PEOPLE SCREENED
(MILLIONS)
SCREENING EXPENDITURES
($ BILLIONS)
Year
Continued
incentives
Increased
incentives
Decreased
incentives
Continued
incentives
Increased
incentives
Decreased
incentives
2010 7.3 7.3 7.3 8.7 8.7 8.7
2035 10.4 14.5 8.3 21.2 29.7 17.0
Cumulative
(2010-2035)
235.2 292.4 206.6 375.6 473.6 326.6
COLORECTAL CANCER
A Markov model was created to assess the effects of improved colorectal cancer screening technology on
treatment and outcomes. The effect of screening on the colorectal cancer PRC as well as the number of
cancer cases prevented through polypectomy was examined for each scenario.
Colorectal cancer originates in polyps, or abnormal growths, in the colon (also known as the large intestine)
or rectum. Not all polyps have the potential to develop into colon cancer, and fewer than 10 percent actually
do. It can take more than 10 years to develop into disease. Once identified, a polyp can be removed through
a polypectomy, preventing a malignancy from occurring. If colorectal cancer develops, patients must go
through treatment, an expensive process that severely affects his or her quality of life.

56 Healthy Savings
Americans are advised to begin colorectal cancer screening at age 50 and repeat at 10-year intervals.
These frequencies are built into the model.Though rare, the disease can occur before 50, and we incorporated
this into our model. Our model includes age-stratified incidence rates from Surveillance, Epidemiology,
and End Results (SEER), a program of the National Cancer Institute (NCI). If a patient is screened and polyps are
detected, normally a polypectomy is performed and a colonoscopy is ordered in three years as surveillance.
Because the time required for a precancerous polyp to progress to cancer varies by the individual, based on a
literature review we estimated that one-third of polyps would do so over 30 years. Screening can also identify
colorectal abnormalities that were not detected through symptom identification. Such procedures can lead
to early diagnosis, which may be represented by a higher probability of detection at an earlier cancer stage.
Among our incentive scenarios, the primary difference is the implications of varied screening rates for
colorectal cancer deaths. The continued incentives scenario assumes the persistence of the annual change in
screening rates derived from MEPS. The annual screening rate increase is doubled in the increased incentives
scenario as advanced screening technologies are deployed, and it is reduced by half in decreasing incentives.
These changes in rates over time are accounted for as changes in the likelihood of the eligible population
actually being screened within the model.
Colorectal cancer PRC was 616,500 in 2010, according to the utilization data in MEPS, which accounts for all
patients with health-care expenditures related to the disease. SEER data reveals a prevalence of 1.2 million
patients, almost double the PRC observed in MEPS. This disparity could be explained by the fact that not all
patients with colorectal cancer have expenditures in the year assessed.
In the continued incentives scenario, the PRC increases from 600,000 to 1.7 million, while under increased
incentives, the PRC increases to 1.4 million. (See Projections: Colorectal Cancer.)The 280,000 reduction in future
PRC in the increased incentives scenario could be caused by the increased screening rate. In the decreased
incentives scenario, the 160,000 additional PRC compared to the continued incentives scenario is consistent
with weaker adherence to screening and therefore less cancer prevention through polypectomy.
From historical trends, it was clear that 2010 expenditures were significantly different from those of previous
years, so to project expenditures we used an average of 2008-2010 per-PRC data.
Colorectal cancer treatment expenditures increase with a rising PRC. The reduction in PRC due to doubling
the increase in screening rates is associated with $19 billion in cumulative savings over 25 years (which
translates to $10.6 billion in 2010 dollars). On the other hand, the expanding PRC associated with decreased
incentives aggregates to $10.7 billion more spending ($6 billion in today’s dollars) than in the continued
incentives scenario.
We further estimated the number of cancer cases prevented, along with associated reductions in the economic
impact, using polypectomy data from HCUP. Since not all polyps will turn into cancer, we assumed that
approximately one-third of polypectomies prevented the disease from developing. Our calculations suggest
that in 2010, 554,400 cases were prevented by screening.
Polypectomy projections from the model show that in 2035, 1.1 million cases will be prevented under the
continued incentives scenario, slightly fewer than under increased incentives (1.2 million). From $12.2 billion
in 2010, the gain to the health-care system and GDP rises tremendously in future years, reaching $45.6 billion
in 2035 in the continued incentives scenario. Compared to that projection, increased incentives generates
additional savings of $90.2 billion over 25 years, or $58.5 billion in 2010 dollars.

There is a chance that colorectal cancer screening incurs costs unrelated to the disease, since the majority
of the screening population is well. Tests can produce mistaken diagnoses, or false positives. MEPS data for
per-PRC expenditures indirectly accounted for false positive treatment costs. However, a potential increase
in such readings as a consequence of screening was not considered.
As mentioned earlier, the expenditures involved in screening the healthy population add a burden to the
health-care system. Our historical estimates show that in 2010, 14.9 million people without colorectal cancer
or precancerous polyps were screened, generating a cost of $17.7 billion. By 2035, this increases to 28.4 million
people and costing $57.9 billion in the continued incentives scenario. A cumulative $893.2 billion, it is
estimated, would be spent on screening the healthy population over 25 years. More frequent screening
in the increased incentives scenario would result in a cumulative $1 trillion in spending, and decreased
incentives channels $838 billion into screening the healthy population.
We also consider the indirect economic effects of colorectal cancer for each scenario. These effects stem
from the employed population with colorectal cancer as well as people saved from having the disease.
Between now and 2035, the indirect impact of colorectal cancer will ease as more advanced screening
technologies are deployed, decreasing the cancer PRC and improving survival, and consequently limiting
productivity loss. By then, the increased incentives scenario would produce a cumulative economic gain
of $198.2 billion compared to the baseline scenario, $23.4 billion of which can be credited to increased
survival. The economic gain would total $109 billion in 2010 dollars. In contrast, the reduced incentives
scenario, involving less investment in technology, would reduce GDP by $112.4 billion compared to the
baseline scenario, or $61.4 billion in 2010 dollars.
Under increased incentives, colorectal cancer prevention through screening boosts GDP by $65 billion
compared to the continued incentives scenario, while the decreased incentives projection results in a loss
to GDP of $53.8 billion.

58 Healthy Savings
PROJECTIONS: COLORECTAL CANCER
: Colorectal cancer treatment and prevention
2010 2015 2020 2025 2030 2035 2010 2015 2020 2025 2030 2035
$ billions$ billions
0
5
10
15
20
25
30
35
40
-25
-20
-15
-10
-5
0
0.00
2.12
5.80
12.08
21.69
34.75 0.00
-1.17
-3.29
-7.25
-13.84
-23.70
Economic impact of colorectal cancer, 2010-2035 ($ billions)†
Due to survival 23.4 -5.6
Total 217.3 -123.2
Projected colorectal cancer population affected, and cases prevented (millions)
PREVENTION
PRC* EPRC** ECC*** NUMBER OF CASES PREVENTED
Year
Continued
incentives
Increased
incentives
Decreased
incentives
Continued
incentives
Increased
incentives
Decreased
incentives
Continued
incentives
Increased
incentives
Decreased
incentives
Continued
incentives
Increased
incentives
Decreased
incentives
2010 0.62 0.62 0.62 0.45 0.45 0.45 0.08 0.08 0.08 0.6 0.6 0.6
2035 1.69 1.41 1.85 1.23 1.03 1.35 0.22 0.18 0.24 1.1 1.2 1.0

Projected economic impact of colorectal cancer ($ billions)†
PREVENTION
TREATMENT EXPENDITURES INDIRECT IMPACT
TREATMENT EXPENDITURES AND
INDIRECT IMPACT
Year
Continued
incentives
Increased
incentives
Decreased
incentives
Continued
incentives
Increased
incentives
Decreased
incentives
Continued
incentives
Increased
incentives
Decreased
incentives
2010 5.4 5.4 5.4 28.1 28.1 28.1 -12.2 -12.2 -12.2
2035 14.7 12.3 16.2 178.1 149.1 195.6 -45.6 -48.9 -40.8
Cumulative
(2010-2035)
317.7 298.7 328.5 2,778.9 2,580.7 2,891.3 -697.0 -787.2 -625.5
In 2010 dollars
(2010-2035)
214.7 204.1 220.7 1,790.5 1,681.5 1,851.9 -452.0 -510.5 -407.7
DIRECTMEDICALEXPENDITURES
HEALTHYPEOPLESCREENED*
(MILLIONS)
SCREENINGEXPENDITURES
($BILLIONS)
Year
Continued
incentives
Increased
incentives
Decreased
incentives
Continued
incentives
Increased
incentives
Decreased
incentives
2010 14.9 14.9 14.9 17.7 17.7 17.7
2035 28.4 35.5 24.8 57.9 72.4 50.7
Cumulative
(2010-2035)
551.9 613.0 521.4 893.2 1,003.4 838.1
* Includes those receiving screening who did not have cancer or were not prevented from developing cancer.

61
TAX REVENUE
AND MEDICAL TECHNOLOGY
I
n this report, we have estimated the effects on GDP due to changes in labor market outcomes associated
with the use of medical devices. Consequently their use also affects the federal personal income tax
revenue generated. For example, if insulin pumps reduce lost workdays and improve productivity for
patients and their caregivers compared to those who inject insulin, this additional value contributed
translates into greater tax revenue. To measure the tax revenue generated by the use of a technology
compared to another or no technology, we estimated a wage-based indirect impact associated with the
technology studied. This approach is similar to that used for GDP-based indirect impact estimates, but we
used average employee wage rather than GDP. The results can be seen in the table below.
Table 8
Wage-based indirect impact associated with medical technology
($ millions)
TECHNOLOGY 2005 2010 AVERAGE (2008-2010)
Insulin pump 493.3 1,003.8 893.8
Heart disease diagnostics and surgery 15,873.8 15,544.7 15,116.0
MRI and joint replacement surgery 6,061.2 5,741.9 6,053.5
Detection 4,888.9 6,758.5 5,624.3
Prevention - -4,254.4 -3,459.8
The average annual (2008-2010) wage-based indirect impact (or the foregone labor income) by insulin pump
users was $893.8 million. It was $15.1 billion and $6.01 billion, respectively, for heart disease and musculoskeletal
disease patients using diagnostics and/or surgery. Similarly, colorectal cancer patients who were screened
had a wage-based indirect impact of $5.6 billion. However, thanks to cases in which cancer was prevented
by screening, $3.4 billion was added to the economy in the form of labor income. Thus, the total indirect
impact of detection and prevention was $2.2 billion.
A portion of this foregone labor income was taxable. In 2010, the median family income in the United States was
$60,23625
and the tax rate for married couples falling within the median income level was 15 percent.26
Applying
that rate historically, we calculated lost tax revenue associated with foregone income estimated from the above
table.The annual average (2008-2010) revenue lost for technology users with these four diseases was $3.6 billion.
25. Current Population Survey, United States Census Bureau.
26. “Federal Individual Tax Rates History,” Tax Foundation.

62 Healthy Savings
If we compare alternative treatments, however, there is actually an income gain associated with using
technology. The additional average annual income generated by pump users (compared to non-users) was
$2,371 per person affected. Similarly, the difference in labor income between people using technology
associated with heart disease and musculoskeletal disease and those who did not amounted to $2,902 and
$12,749, respectively. Patients screened for colorectal cancer and their caregivers earned $43,194 more than
those not screened. Further, colorectal cancer screening prevented individuals from developing cancer,
which would bring in an extra $20,276 per case.
Table 9
Difference in wage-based indirect impact associated with technology
Per person affected, compared to non-users ($)
Insulin pump 1,963.7 2,104.5 2,371.2
Heart disease diagnostics and surgery 5,548.2 3,067.7 2,902.3
Detection 16,668.0 44,279.6 43,194.4
Prevention - 19,541.2 20,276.4
With medical devices/technology strengthening the labor market, these income gains generate tax revenue
and expand the economy. Using a constant 15 percent tax rate, the additional revenue generated by insulin
pump users is $356 per person affected. Tax revenue generated by heart disease and musculoskeletal
disease patients who use technology is $435 and $1,912 per person, respectively. Similarly, additional tax
revenue is $6,479 for colorectal cancer patients who were screened. Such screening also produces $3,041
in tax revenue per person affected due to prevention.
Table 10
Tax revenue generated by medical technology users
Per person affected, compared to non-users ($)
Insulin pump 294.6 315.7 355.7
Heart disease diagnostics and surgery 832.2 460.2 435.3
Detection 2,500.2 6,641.9 6,479.2
Prevention - 2,931.2 3,041.5

63
MAIN TAKEAWAYS
A
s sedentary ways of life and unhealthy eating habits take their toll, severe ailments such as diabetes,
cancer, and heart and musculoskeletal disease are likely to flourish among America’s aging populace.
We are already seeing evidence of that. While the risk spreads, however, medical technology can
play a crucial role in prevention, early detection, and better management of disease.
We studied a group of technologies that have proved their effectiveness for these purposes. Our work
suggests that routine measures such as colonoscopy or sigmoidoscopy might have prevented 560,000
cases of colorectal cancer annually from 2008 to 2010. Further, if we follow the continued incentives
scenario into the future, about 1.1 million cases of the potentially lethal disease will be prevented in 2035.
The same technology is vital to early detection efforts. Early detection of a malady improves a patient’s
chance of survival and may make him or her eligible for less invasive and less disruptive treatment. If heart
disease, for instance, is diagnosed in its initial stages, surgery may be unnecessary and medicine the better
option. Finally, after the onset of a chronic disease, it must be managed well to afford the best quality of
life possible for the patient. Insulin pumps have been found more effective than injections in managing
adverse effects for diabetics, such as insulin spikes.
These technologies have been criticized for the costs involved in needless testing of healthy populations.
Some say their widespread use has been draining the health-care system. In this study, aggregate screening
expenditures on healthy people were $31 billion annually from 2008 to 2010. Annual expenses for patients
using the studied technologies were $51.6 billion higher than those for non-users.
However, there are powerful benefits to consider. Due to more effective disease management, it is possible
that the more expensive treatments and sites of service can be avoided, yielding savings across the system.
Additionally, by extending survival in many cases and improving quality of life, these medical technologies aid
patients’ability to work and labor market outcomes overall. For patients and their informal caregivers as well,
fewer workdays are lost and productivity is enhanced. Indeed, during the 2008-2010 period, these factors led
to an average annual GDP gain of $106.2 billion and increased federal tax revenue by $7.2 billion.
In our view, these effects are likely to fortify future GDP growth, job creation, incomes, and government
revenue. In other words, there is a worthy economic rationale for investing in medical technology,
if strengthening our arsenal against chronic disease is not compelling enough.
There is a worthy economic rationale for investing in
medical technology, along with waging the battle for
better health and longer lifespans.

65
METHODOLOGY
Technology and the Economic Burden of Disease: Historical Trends
This report uses a cost-of-illness approach to estimate the trends from 2005 to 2010 associated with
treatment of the studied diseases (diabetes, heart disease, musculoskeletal disease, and colorectal cancer).
“Economic burden” is the aggregate of direct treatment expenditures, indirect economic impact (in terms
of foregone gross domestic product), and expenditures associated with screening the healthy population.
The benefit or loss of using the technology is measured as the difference between the economic burden of
using technology to treat a disease and the economic burden associated with not doing so.
Treatment expenditures
The cost-of-illness approach represents actual treatment expenditures and reflects the range of treatment
options and costs to patients. The information was compiled by site of services, which includes treatment/
procedures performed by health-care professionals in their offices, outpatient services, inpatient hospital
care, emergency rooms, prescription drug expenditures, and home health care. Actually, prescription drug
expenditures and home health care should be considered product categories, but they were included in
the site of services category to simplify the discussion. Our framework links the disease-related number of
patients, or population reporting a condition, and site-specific treatment expenditures.
Data sources
Medical Expenditure Panel Survey
Disease-related treatment expenditures and population reporting a condition (PRC) data is obtained from
the Medical Expenditure Panel Survey (MEPS) collected by the Agency for Health Care Research and Quality
(AHRQ), a unit of the U.S. Department of Health and Human Services. MEPS is a nationally representative
sample of the noninstitutionalized civilian population with annual data on the provision of health services,
site of service, frequency, and associated payment. MEPS was designed to provide policymakers, health-care
administrators, businesses, and others with timely, comprehensive information about health-care use and
costs in the United States.
As such, MEPS is unparalleled for the depth of its data and links. Since the current release of data is
comparable to that of earlier medical expenditure surveys, it is possible to analyze long-term trends in
disease treatment costs. MEPS is a large-scale survey of families, individuals, and their medical providers
across the U.S., collecting data on individuals’use of services and the associated costs.
The MEPS database has two major parts: a household component (HC) and an insurance component. It also
includes a supplemental medical provider component (MPC) and a nursing home component (available
only for 1996). The HC is particularly relevant to this study because it draws upon a nationally representative
subsample of households that participated in the prior year’s National Health Interview Survey (NHIS).
Public-use data in the HC contains demographic characteristics, health conditions, health status, and use of
medical services for more than 30,000 people each year. Individual data can be used to make estimates for
the noninstitutionalized civilian population by using population-based weighted factors.

66 Healthy Savings
MEPS’HC public-use data files consist of consolidated full-year data and medical event files. A person-level
consolidated data file provides expenditure and utilization data for the calendar year from several rounds of
collections. Medical event files provide calendar year information on unique household-reported medical
events. They consist of seven individual files characterized by site of service: office-based medical provider
visits, hospital outpatient visits, inpatient hospital care, emergency room visits, prescribed medicines, home
health care, dental visits, and other medical expenses. For the purposes of this study, dental visits and other
medical expenses were not included. Person-level expenditures associated with a disease type are derived
and aggregated from these individual data files.
For disease information, MEPS data provides both three-digit International Classification of Disease, 9th Revision,
Clinical Modification (ICD-9) codes and Clinical Classification Software (CCS) codes. CCS codes were
generated by grouping ICD-9 codes into 260 mutually exclusive, clinically meaningful disease categories.
Many chronic diseases of interest for this analysis are included in these categories, such as diabetes and
heart disease, while other chronic diseases were aggregated from multiple categories, such as colorectal
cancer and musculoskeletal disease.
The MEPS database uses CCS and ICD-9 codes to indicate the conditions for which each patient is treated.
Some conditions have related survey questions in MEPS recording whether each respondent had ever
been diagnosed with one of the assessed conditions by a health-care provider. There may be discrepancies
between this question and the condition codes; only condition codes were used for consistency.
MEPS uses ICD-9 codes to identify expenditures and medical visits related to procedures associated with
some of the examined medical technology. All conditions and procedures are self-reported in the interview
process that obtains data for MEPS. Because procedures are not prompted for in the interview process,
technology use is generally underreported.
Please refer to Table A15 for details on the CCS and ICD-9 codes used in this study.
AHRQ provides useful national-level MEPS summary data tables on expenditures and population reporting
a condition for 60 selected chronic conditions (such as heart disease and diabetes), which can be used to
benchmark our estimates.
Healthcare Cost and Utilization Project
The Healthcare Cost and Utilization Project (HCUP), sponsored by AHRQ, represents one of the largest
national hospital databases. HCUP has several databases: Nationwide Inpatient Sample, Kids’Inpatient
Database, Nationwide Emergency Department Sample, State Inpatient Databases, State Ambulatory
Surgery Databases, and State Emergency Department Databases. For the purposes of this study, we used
the Nationwide Inpatient Sample (NIS) 27
because it is the most comprehensive for the procedures that are
included, such as knee and hip replacements. NIS is the largest publicly available all-payer inpatient care
database in the U.S. and contains data from approximately 8 million hospital stays each year. More than
1,000 community hospitals (which exclude long-term care hospitals, federal hospitals, etc.) are included in
the database from approximately 45 states. NIS produces nationally representative figures, with hospital
discharges as the main variable.
27. Nationwide Inpatient Sample, HCUPnet.

67Methodology
For disease information, NIS data provides CCS and ICD-9 codes. ICD-9 coding is widely used by health-care
professionals to classify both diseases and procedures. NIS categorizes principal diagnosis and all-listed
diagnosis by ICD-9 and CCS codes. A principal diagnosis is the reason for admission, while an all-listed
diagnosis can include the principal plus additional conditions present at the time of admission or that
develop during the hospital stay and affect treatment.
For procedure information, NIS data also provides CCS and ICD-9 codes by principal procedure and all-listed
procedure. Similar to diagnosis data, all-listed procedures include all of those performed during the hospital
stay, while principal procedures are those undertaken for definitive treatment, rather than exploratory or
diagnostic purposes.
NIS calculates cost data based on a conversion method using cost-to-charge ratios taken from hospital
accounting reports from the Centers for Medicare and Medicaid Services (CMS). In some cases, only charge
data is available, which is generally higher than cost as it represents the amount hospitals charge for services.
Data calculation
Utilizing MEPS individual event data files associated with each site of service, we estimate expenditures and
PRC for all the examined diseases: diabetes, heart disease, musculoskeletal disease, and colorectal cancer.
Treatment expenditures of four analyzed chronic diseases were assessed through analysis of the MEPS
database from 2005 to 2010. For each disease, health-care expenditures associated with having the condition
and using the predetermined medical technologies were calculated on an annual basis. Appropriate weights
(as specified in MEPS) were used to calculate nationally representative figures.
Expenditures were calculated for all office-based, outpatient, inpatient, emergency room, prescription
drug, and home health care. Expenditures rather than charges were used to ensure that all costs levied on
the health-care system were included. Disease-related expenditures were calculated as all expenditures
of visits associated with the relevant condition codes. Technology-related expenditures were calculated
as all expenditures of visits associated with the relevant condition code if the specific patient used
the technology at a disease-related visit during that calendar year. For example, this would include all
musculoskeletal disease-related expenditures in a given year for a patient receiving a knee replacement.
PRC is the number of unique patients with visits associated with a condition at any site of service who
incurred an expenditure. Patients were designated as having undergone a procedure in a certain year if they
had a disease-related health-care visit attached to an ICD-9 procedure code. After expenditures and PRC were
calculated for each disease and disease-related medical technology, disease-specific annual expenditures per
PRC were calculated for each site of service. Expenditures per PRC can be perceived as the average treatment
expenditure for each disease.
Data adjustment
A problematic aspect of MEPS data is the variation across years. Because of this, trends were assessed
chronologically, and outliers were smoothed by calculation of three-year averages. Expenditures and counts
based on a small original sample size (fewer than 10 patients) were eliminated, and total expenditures and
PRCs were updated accordingly. Below, we discuss methodologies associated with diseases and technology
that were not directly available in MEPS data.

68 Healthy Savings
Diabetes
The MEPS survey collects information about how many diabetes patients use insulin but does not distinguish
by the mode of administration. To estimate the PRC using insulin pumps and associated expenditures,
MEPS data was used in conjunction with relevant literature. According to Bode et al. in a presentation on
Medtronic, approximately 7 percent of insulin users used pumps in 2010.28
According to MEPS, there were
about 5.3 million insulin-dependent PRC in 2010, with slightly more than 379,000 of them pump users.
In a previous paper, Bode et al. showed insulin pumps use from 1997 to 2000, which was combined with
CDC data on the insulin user population to determine a trend in pump use.29
That was used to estimate the
proportion of pump users from 2005 to 2009.
Figure A1
Proportion of insulin dependent diabetes patients using a pump
Percent
0
1
2
3
4
5
6
7
8
2000
2.7
3.1
3.5
4.0
4.4
4.8
5.3
5.7
6.2
6.6
7.1
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
The PRC associated with sites of service was determined separately from PRC for insulin pump usage.
Initially the ratio of pump users to total insulin users was applied across all sites of service to get a base
number. Schuffham and Carr30
reported that using insulin pumps is associated with fewer hypoglycemic
events that might require inpatient hospital stays or emergency room visits. The study showed that
inpatient admission falls 43 percent for insulin pumps users and ER visits fall 53 percent.
Prescription-related PRC was assumed to be equal to the total PRC since insulin use falls under prescribed
medication.SincefewinsulinusersmaynotincurRXexpendituresforthatyear,thismethodologyprovidesanupper
bound. Because no specific data was found to inform a reduction in office-based and outpatient PRC, we assumed
a 35 percent reduction in office-based and outpatient services used by patients based on the assumption that the
insulin pump reduces inpatient expenditures by $0.62 for every $1 of such expenditures for other insulin users.
Using this logic, inpatient expenditures were assumed to be 38 percent of the base expenditure per PRC.
Emergency room expenditures were assumed to be 21 percent, comparable to the standardized coefficients
28. Bode et al. “Insulin Pump presentation for Medtronic.”
29. Bode et al. “Diabetes Management in the New Millennium using Insulin Pump Therapy.” Diabetes/Metabolism Research and Reviews 18:1 (2002) p.
S14-S20.
30. Scuffham and Carr, “The Cost-Effectiveness of Continuous Subcutaneous Insulin Infusion Compared with Multiple Daily Injections for the
Management of Diabetes,” Diabetes Medicine 7 (2003) p.586-93.

69Methodology
for inpatient expenditures from the same study. Since no data was found to inform changes in office-based,
outpatient, and home health expenditures per PRC, a 50 percent reduction was assumed based on reduced
use of clinician time and decreased use of tests associated with better controlled diabetes.
Heart disease
If a patient received care attributed to the ICD-9 procedure code for angioplasty in 2010, that individual
would be identified as having an angioplasty for that year. The heart disease-related utilization
and expenditures for these patients were then calculated for each site of service. All utilization and
expenditures for heart disease were recorded for sonogram, electrocardiogram, and x-ray patients to
calculate expenditures and PRC for this subset of patients.
Musculoskeletal disease
Similar methods were followed for joint replacement surgery and musculoskeletal disease-related utilization
and expenditures. The use of medical technology such as MRI or joint replacement surgery was recorded for
each visit in the medical event file. Patients were identified as using one of these medical technologies if at
least one disease-related health-care visit indicated use in the assessed calendar year. Therefore, if a patient
underwent an MRI at any site of service with an expenditure attributed to musculoskeletal disease, then that
patient would be identified as having an MRI that year. All utilization and expenditures for musculoskeletal
disease would be recorded for MRI patients to calculate expenditures and PRC for this subset of patients.
Colorectal cancer
The variable referring to screening rate was linked to the full-year files in MEPS. The 2005-2008 MEPS
questionnaires recorded whether screening for colorectal cancer was performed in the past five years.
The 2009 and 2010 questionnaires recorded the preceding question and additionally asked respondents
whether a colonoscopy was performed in the preceding 10 years and flex sigmoidoscopy in the preceding
five years, in accordance with current screening guidelines. To overcome data differences across years,
2005-2008 survey year data were benchmarked to 2009-2010 data.
Colonoscopy/sigmoidoscopy plays an important role in disease prevention. This screening technology can
locate polyps and remove them through a procedure called polypectomy. If left alone, some polyps will
develop into cancer. This study incorporates the number of cases prevented and expenditures avoided
through screening and early detection. To estimate the number of cases prevented from developing into
colorectal cancer, estimates from HCUP were applied to findings from MEPS. First, the number of screenings
(estimated as number of colonoscopies) of both cancerous and healthy patients was calculated by multiplying
the number of cancerous patients with a colonoscopy in MEPS with the ratio of all colonoscopies performed
to colonoscopy performed on colorectal cancer patients from HCUP.
Next, after calculating the percentage of screened individuals undergoing polypectomy and separating
those from cancer patients from HCUP, we estimated the MEPS equivalent of the total number of cancerous
patients with a polypectomy. Since not all polyps will become cancer, we assumed that one-third of
non-cancerous polypectomies would have resulted in cancer to determine the historical number of
cases prevented. Using similar methodology, we estimated the associated avoided expenditure per case
prevented. The total amount saved for the health-care system is obtained by multiplying the number of
cases prevented with the expenditure per person.

70 Healthy Savings
Historical indirect impact
Good health can largely determine a working person’s economic contribution. When individuals suffer from
chronic disease, the result is often diminished productivity in addition to lost workdays, or absenteeism.
An ill employee who shows up for work—to avoid taking sick days, for example—may not perform well,
a circumstance known as presenteeism.
Informal caregivers also contribute to lost productivity through missed workdays and presenteeism.
Currently, more than 20 million full-time employees provide informal care to others.31
For this study,
therefore, it is necessary to consider both employee groups for a more complete picture of the indirect
impact of both chronic disease and its associated technology due to absenteeism and presenteeism.
Our calculation of indirect impact measures labor market outcomes related to work loss and productivity.
First, any individual suffering or who has suffered from a chronic disease will have two main effects on work,
absenteeism and presenteeism. Similarly, any person taking care of individuals with chronic disease will see
an adverse impact on his or her work. Hence, the indirect impact of both overall disease and the effects of
technology is the aggregate value (in terms of foregone GDP) of the following:
1) Indirect impact due to individual’s absenteeism
2) Indirect impact due to individual’s presenteeism
3) Indirect impact due to caregiver’s absenteeism
4) Indirect impact due to caregiver’s presenteeism
Data sources
National Health Interview Survey
Information about individual absenteeism was mainly obtained from the National Health Interview Survey
(NHIS), a representative sample that asks various health-related questions regarding conditions, employment,
treatment, and cancer screening. The NHIS has several components including the family core, a household
level, person level, a sample adult file, sample adult cancer file, and a sample child file. For the purposes of this
study, we relied primarily on the sample adult and sample adult cancer files. These two files are representative
of the adult U.S. population when appropriately weighted.
Since the NHIS does not ask about the number of lost workdays for a particular disease, we used a proxy
measurement. For example, one survey question from the sample adult file asks, “During the past 12 months,
about how many days did you miss work at a job or business because of illness or injury (do not include
maternity leave)?”We matched all employed individuals who has had a particular chronic disease (the
employed population reporting a condition, or EPRC) with the number of lost workdays in the past 12 months
due to illness or injury. Individuals with borderline disease are not included as part of the PRC or EPRC.
We used this method to derive the number of lost workdays for each disease.
To overcome historical variations, outliers were identified and adjusted. The indirect impact in this study is
estimated on the basis of lost employee output, or foregone GDP (except the special case of lost wages and
31. “Caregiving in the U.S. Executive Summary,” National Alliance for Caregiving, AARP, 2009.

71Methodology
changes in tax revenue from technology use), to capture the full impact on the economy as a whole. Lost
wages and changes in tax revenue associated with disease and technology use are examined separately. To
estimate individual absenteeism, we multiplied the number of lost workdays to a daily average of GDP to
the daily equivalent per employed person.
Once we estimated indirect impact of an individual’s lost workdays, we followed a 2004 study by Goetzel
et al32
to estimate an individual’s (EPRC) presenteeism. The study reported costs related to absenteeism and
presenteeism (in addition to treatment costs) by disease. The following table summarizes the findings from
the Goetzel study.
Table A1
Costs of absenteeism and presenteeism
$ per employee, annual
CHRONIC DISEASE ABSENTEEISM PRESENTEEISM
Diabetes 19.24 158.75
Heart disease 19.21 70.53
Musculoskeletal disease* 15.54 143.11
Any cancer 4.46 75.71
*Since the study provides estimates for arthritis, a lower presenteeism estimate was used to represent the broader
category of musculoskeletal disease.
We use disease-specific ratios of presenteeism to absenteeism (from the Goetzel study) and our estimates
from individual lost workdays to derive indirect impact due to individual presenteeism. The Goetzel study
provides the value of absenteeism and presenteeism for any cancer. To determine the difference between
absenteeism and presenteeism associated with colorectal cancer, we used an adjustment model linked to
the five-year survival rate by cancer type.
Table A2
5-year survival probabilities by type of cancer
SITE SURVIVAL (%)
All cancer sites (invasive) 68.1
Breast 89.2
Colon and rectum 64.9
Lung and bronchus 16.6
Prostate 99.2
Source: National Cancer Institute.
32. Goetzel et al., “Health, Absence, Disability, and Presenteeism Cost Estimates of Certain Physical and Mental Health Conditions Affecting U.S.
Employers,” Journal of Occupational and Environmental Medicine 46, (2004).

72 Healthy Savings
Using adjustments, the ratio between absenteeism and presenteeism associated with overall colorectal
cancer was estimated at 16.2.
Indirect impact associated with caregivers
To estimate the impact of lost caregiver workdays, we first use estimates from two studies. In 2004, the National
Alliance for Caregiving and AARP33
reported that there were about 21.5 million full-time employed caregivers.
In 2009, an updated study revealed that 22.5 million (46 percent) of the 48.9 million caregivers (to adult
recipients) were employed full-time.34
Comparing the growth of full-time employed caregivers between 2004
and 2009, we assumed an increase of 200,000 caregivers per year.
The second study,35
conducted by Metlife, estimates that 10 percent of male caregivers miss, on average,
nine workdays a year. Among female caregivers, 18 percent miss an average of 24.75 workdays. Caregivers’lost
workdays were estimated using the above information for 2003 and 2004. Estimates for all other years were
calculated in proportion to the 2003 absence days for all caregivers to the number of full-time employed
caregivers. Caregivers’absenteeism for each disease is calculated in a similar manner to individual absenteeism.
To estimate caregiver presenteeism, we first calculated the number of employed caregivers by condition,
or ECC. This is estimated by multiplying the total number of full-time caregivers by the ratio of individual
EPRC to national employment. Caregiver presenteeism is then calculated similar to that of individual
presenteeism. Following a study by Levy,36
we allocated 75 percent of ECC-adjusted individual presenteeism
as caregiver presenteeism. We may further adjust caregiver presenteeism by disease, as assumed below.
Wage-based indirect impact
Changes in labor market outcomes associated with the use of medical devices also affect the federal
personal income tax revenue generated. For example, if insulin pump use reduces lost workdays and
improves productivity for patients and their caregivers compared to those who inject insulin, this additional
value contributed translates into greater tax revenue. We calculated the changes in tax revenue to further
assess the effects of medical technology on the economy.
Similar to the methodology in the GDP-based approach, we used a daily average wage rate to calculate
the wage-based indirect impact separated by disease and technology use for individuals and caregivers.
In 2010, the median family income in the United States was $60,236,37
and the federal marginal tax rate for
married couples falling within the median income level was 15 percent.38
Using this rate as a constant for all
years, we calculated the tax revenue added or lost to the economy. The difference in tax revenue generated
from those who used technology compared to those who did not represents a tax gain/loss.
33. National Alliance for Caregiving and AARP: “Caregiving in the U.S.,” 2004.
34. “Caregiving in the U.S. Executive Summary,” National Alliance for Caregiving, AARP, 2009.
35. Metlife Mature Market Institute, National Alliance for Caregiving, 2006, “The Metlife Caregiving Cost Study: Productivity Losses to U.S. Business”.
36. David Levy, “Presenteeism: A Method for Assessing the Extent of Family Caregivers in the Workplace,” (American Association for Caregiver
Education, 2003). See also: David Levy, “Presenteeism: A Method for Assessing the Extent of Family Caregivers in the Workplace and Their
Financial Impact,” (American Association for Caregiver Education, 2007).
37. Current Population Survey, United States Census Bureau.
38. “Federal Individual Tax Rates History,” Tax Foundation.

73Methodology
The following sections provide a detailed methodology of how we measured disease-specific indirect
impact associated with medical technology.
Diabetes
In order to measure the indirect impact of diabetic patients who use insulin pumps, we first measured
indirect impact for all insulin users. The NHIS asked respondents, “Are you now taking insulin?” We used this
question in conjunction with questions regarding lost workdays due to illness and PRC to calculate EPRC
and lost workdays for diabetic patients and also those who depend on insulin.
From the estimates of treatment expenditures, the proportion of insulin dependent diabetics using an insulin
pump was used to estimate the insulin pump EPRC. We assumed that insulin pumps lower absenteeism and
presenteeism, owing to better disease management. Using a study by Scuffham and Carr39
that says pump
use reduces hypoglycemic events by 13 percent, we adjusted lost workdays accordingly. The insulin pump
also reduces the number of complications, such as blindness, nerve damage, and renal disease, but hypoglycemic
events were used for a conservative estimate because they represent a short-term measure that can
potentially affect any diabetic using insulin.
Using the estimated EPRC and lost workdays for insulin pump users, absenteeism was calculated in a manner
similar to overall diabetes and all insulin users. Reducing diabetic complications such as hypoglycemic
events often makes patients feel less anxious and their quality of life improves, reducing presenteeism as
well. Research shows that the quality of life for those who inject insulin is 5.3 percent worse than those
using pumps.40
This ratio was applied to determine the reduction in presenteeism for insulin pumps users
compared to those who inject. The methodology for caregivers followed a similar approach.
Heart disease
NHIS asks respondents whether they have ever been told by a doctor or health professional that they
had coronary heart disease, angina, a heart attack (myocardial infarction), or other kind of heart condition.
All categories of heart disease are aggregated to estimate EPRC and related indirect impact. We use heart
disease surgery as a proxy to determine the indirect impact of both diagnostic and surgical technology on
productivity. Again, this is because surgery has a larger effect on productivity than a diagnostic test.
We use the general surgery question from NHIS in combination with people who have heart disease to
calculate EPRC and work loss days for heart disease patients who had surgery in the past year. To ensure
that the EPRC captured only that category of patient, we used the percentage of heart disease PRC who
had a heart disease procedure from the treatment expenditure calculations. Associated work loss days were
calculated using information from Abbas et al.41
On average, 51 percent of heart attack patients can expect
to return to work within one month, and 78 percent return to work after six months. With these facts,
we adjusted lost workdays accordingly. Using EPRC and lost workdays for procedures specific to heart
disease, absenteeism was calculated in a similar manner to that for overall heart disease.
39. P. Scuffham and L. Carr, “The Cost-effectiveness of continuous subcutaneous insulin infusion compared with multiple daily injections for the
management of diabetes,” Diabetic Medicine 20 (2003), pp. 586-593.
40. Ibid.
41. Amr E. Abbas et al., “Frequency of Returning to Work One and Six Months Following Percutaneous Coronary Intervention for Acute Myocardial
Infarction,” American Journal of Cardiology 94 (2004).

74 Healthy Savings
Technology can also raise quality of life for heart disease patients and reduce presenteeism. According to Rosen
et al.,42
surgical revascularization represents a potential 22.4 percent quality of life increase if it prevents a major
cardiac event. Presenteeism was adjusted accordingly using this information. Since ECC and lost caregiver
workdays are proportional to individuals, the methodology for caregivers follows similar adjustments.
NHIS asks respondents if they have ever been told by a doctor or health professional that they have some
form of arthritis, rheumatoid arthritis, gout, lupus, or fibromyalgia. Using this question, we first calculated
EPRC and lost workdays for the overall disease.
To calculate EPRC associated with technology, we used surgery as a proxy for technology use because surgery
would affect both absenteeism and presenteeism more dramatically than use of a diagnostic technology.
However, if a diagnostic technology allowed earlier diagnosis and more effective treatment, it would reduce
absenteeism and presenteeism as well. This serves as a conservative estimate in this regard. NHIS also asks,
“During the past 12 months, have you had surgery or other surgical procedures either as an inpatient or
outpatient?” This question was matched up with those who have musculoskeletal disease to estimate EPRC
and lost workdays associated with surgery. To adjust for only joint replacement surgeries, the percentage of
musculoskeletal disease PRC with joint replacement calculated from MEPS was used.
To calculate lost workdays for musculoskeletal disease patients related to musculoskeletal procedures,
we used the information from a study saving that 94 percent of hip replacement patients return to work
within two months and the remaining 6 percent return within a year.43
Assuming that the 94 percent
were out of work for two months and the other 6 percent 11 months, we adjusted accordingly. Using the
calculated EPRC and lost workdays for musculoskeletal disease patients who had related procedures,
absenteeism was calculated in a similar manner to overall musculoskeletal disease.
Overall presenteeism was calculated using ratios from the previously mentioned Goetzel study. However,
technology can improve quality of life for patients with musculoskeletal disease, thereby lowering presenteeism.
David Ruiz et al. estimated that knee replacement surgery added 3.4 QALYs among patients ages 40-44.44
We used this estimate as a proxy for the effects of all musculoskeletal disease-related technology. With ECC
and caregiver lost workdays proportional to EPRC and individual lost workdays, the methodology for
caregivers reflects similar adjustments.
Colorectal cancer
Early detection through screening
NHIS asks respondents separate questions if they have ever been diagnosed with colon cancer and rectal
cancer. EPRC and lost workdays associated with colorectal cancer were created aggregating these questions
together. Absenteeism and presenteeism are estimated consistent with the methodology for overall disease.
42. Virginia M. Rosen et al., “Cost Effectiveness of Intensive Lipid-Lowering Treatment for Patients with Congestive Heart Failure and Coronary Heart
Disease in the U.S.,” Pharmacoeconomics 28, no. 1 (2010).
43. Ryan M. Nunley, et al., “Do Patients Return to Work After Hip Arthroplasty Surgery?” Journal of Arthroplasty 26, no. 6 Suppl. 1 (2011).
44. David Ruiz et al., “The Direct and Indirect Costs to Society of Treatment for End-Stage Knee Osteoarthritis,” Journal of Bone and Joint Surgery 95
(2013), pp. 1473-1480.

75Methodology
This analysis also estimated EPRC and lost workdays associated with cancer patients who had a colonoscopy
in the past 10 years or a sigmoidoscopy in the past five years, which was used as a benchmark to estimate
other years’ EPRC and lost workdays.
For colorectal cancer patients with colonoscopy, EPRC was estimated using the percent of colorectal cancer
PRC who had a colonoscopy.
In order to determine presenteeism gained by early detection, we used utility gains associated with early
stage detection. From Heitman et al. 2010,45
we have colorectal cancer stage distributions for patients
identified through screened and unscreened patients developing colorectal cancer. As to be expected,
a larger proportion of unscreened patients have later stage cancer. Each stage is associated with a utility
mentioned in the Ness et al. study,46
and utilities decrease as the cancer becomes more severe (or increases
in stage). Multiplying the stage utility by the stage proportion for both unscreened and screened cancer
patients yields an average utility. Comparing the average utility for unscreened and screened colorectal
cancer patients reveals an 11 percent increase in utility for screened cancer patients. We assumed that
screening would lower presenteeism by 11 percent as compared to overall cancer patients.
For ECC, lost workdays and presenteeism for caregivers are proportional to those for individuals, so the
methodology for caregivers follows similar adjustments.
Table A3
Utilities associated with screen-detected and symptom-detected colorectal cancer 47
STAGE
PROPORTION OF
UNSCREENED
PROPORTION OF
SCREENED
STAGE
UTILITY
UNSCREENED
WEIGHTED UTILITY
SCREENED
WEIGHTED UTILITY
1 0.15 0.43 0.74 0.11 0.31
2 0.36 0.23 0.74 0.26 0.17
3 0.28 0.27 0.67 0.19 0.18
4 0.22 0.08 0.25 0.05 0.02
Weighted mean 0.61 0.68
Sources: American Journal of Gastroenterology, PLOS Medicine.
Prevention through screening
Using the ratio of PRC for colorectal cancer patients with colonoscopy to non-colorectal cancer patients with
colonoscopy, we estimated employed people prevented from colorectal cancer (EPPCC). Lost workdays were
adjusted to EPPCC. Associated absenteeism prevented and presenteeism prevented both for individuals and
caregivers were consistent with the methodology used for colorectal cancer patients.
45. Steven J, Heitman et al., “Colorectal Cancer Screening for Average-Risk North Americans: An Economic Evaluation,” PLOS Medicine 7, no. 11 (2010).
46. Reid M. Ness et al., “Utility Valuations for Outcome States of Colorectal Cancer,” The American Journal of Gastroenterology 94, no. 6 (1999).
47. Comparing the aggregate utilities associated with screen-detected and symptom-detected colorectal cancer reveals an 11 percent increase in
utility for screen-detected cancer compared to symptom-detected cancer.

76 Healthy Savings
Economic Impact Projections Associated with Medical Technology
This report projected economic impact associated with specific medical device/technology used through
2035. One of the main objectives of the projections was to incorporate future effects of medical device/
technology innovations on disease-specific PRC and expenditures. With this purpose, we simulated three
future scenarios:
1) Continued incentives (baseline) scenario: In this scenario, the growth in medical innovation remains
at the same historical pace, along with the growth rate of its use.
2) Increased Incentives (optimistic) scenario: Medical innovation advances at a higher than historical rate.
3) Decreased Incentives (pessimistic) scenario: Medical innovation progresses at a lower than historical rate.
Projection of PRC
To estimate future treatment expenditures and indirect impact, we first projected the PRC and integrated
other relevant data. An appropriate model for projection of treatment expenditures associated with
disease-specific technology involves a range of stages and options. We used decision trees that illustrate
health processes over time to create disease-specific Markov models.
Markov models
Markov models allow the evaluation of health processes over time. They are created by identifying various
stages of a disease. Each disease includes several probabilities, each of which represent a transition from
one stage to another. Information about these probabilities is obtained from systematic literature review,
public-use health related data sets, and any prior estimated measure (example: historical expenditures per
PRC). This can allow forecasting of disease-specific expenditures and population reporting conditions (PRC).
All models created for this study begin in 2010 and have a cycle length of one year, so a hypothetical
individual in the model can jump to another stage at the end of one year. At each stage, an individual can
transition into the “dead” state, which includes dying of the disease and other causes.48
For each projected
year and each scenario, the PRC for each health state is collected from the model. A cost per PRC derived
from the historical analysis of MEPS is assigned to each health state and overall costs of the disease can
then be calculated. The difference between scenarios in economic impact shows the benefits and losses
associated with investing in technology innovation.
TreeAge Pro Suite 2013 was the software used for all Markov model related analysis. Specific numerical
inputs for variables used in the models can be found in Tables A11 through A14.
Diabetes
We used Markov models to estimate future PRC associated with non-insulin dependent diabetes, insulin
dependent diabetes using insulin injections, and insulin dependent diabetes using insulin pumps.
Combining the PRC for all categories of diabetes reveals the aggregate of insulin pump use on mortality
and expenditures.
48. Data on mortality rates was obtained from the Centers for Disease Control and Prevention (CDC).

77Methodology
The overall model for diabetes had five health states: “No diabetes”; “Non-insulin dependent diabetes“;
“Insulin dependent diabetes using insulin injections“; “Insulin dependent diabetes using insulin pumps“;
and ”Dead.” Generally, as diabetes worsens in a patient, the patient transitions from non-insulin dependent
diabetes to insulin dependent diabetes. Many diabetes patients, including most juvenile diabetes patients,
can require insulin at onset. Currently, the two main methods for insulin delivery are insulin injections and
the insulin pump. While the purchase and maintenance of insulin pumps can be more expensive than
insulin injections, the insulin pump has been shown to improve both health outcomes, especially those
related to hypoglycemic and hyperglycemic events that require more expensive care, and quality of life.
We calculated the percentage of the “No diabetes”well population in 2010 from the MEPS 2010 data.
We assumed that at any time period, an individual can die (we used all-cause mortality rates determined
by the CDC) or develop diabetes, which depends on obesity and high cholesterol. A proportion of new
diabetics will be insulin dependent at onset, a value set to one-fourth of the percentage of diabetics
using insulin in 2010. The probability of new onset insulin dependent diabetics using the pump versus
insulin injections is proportional to that probability in current insulin users in 2010. For more details on
probabilities used and sources, please refer to Table A11.
We used MEPS historic data to estimate the probability of transitioning from non-insulin dependent to
insulin dependent diabetes and further to using injections or a pump. We assumed once individuals were
using an insulin pump, they would not discontinue usage of this technology and would continue to remain
in that state or die.
Individuals in all diabetes states can survive or die based on the CDC death rate multiplied by the relative risk
of death due to diabetes. The death rate was higher for people with diabetes compared to people without
diabetes, and was highest for people on insulin compared to those who were non-insulin dependent
because insulin dependent diabetes is more severe. Insulin pumps were assumed to improve mortality by
20 percent due to decreased risk of hypoglycemic events and long-term complications.
After the initial conditions and transition probabilities are included in the model, a Markov cohort analysis
is performed. This analysis calculated the PRC in each health state for 25 years. Estimated PRCs from the
model were then adjusted to match MEPS 2010 diabetes PRC.
The following diagram shows a simplified version of the model described above for diabetes. A more
detailed version of the same model is included later in Figures A6 through A10.
Figure A2
State transition diagram for diabetes Markov model
Well
Non-insulin
dependent
Insulin
injections
Insulin
pump

78 Healthy Savings
Next, we projected three scenarios involving different levels of incentives for innovations in medical device/
technology. Three separate Markov models for each scenario were specified. Markov models for diabetes
associated with each scenario have the same underlying structure but certain variables were modified.
For diabetes, each incentives scenario will have different likelihoods that people will begin to use the
insulin pump.
Continued incentives
PRC for the “continued incentives” used all variables mentioned above, and assumed a continuation of the
historic rate of insulin pump adoption, which affects the death rate and glycemic episodes.
Increased incentives
PRC for the“increased incentives”scenario assumed a higher take up rate for the insulin pumps (assuming
twice the growth rate of continued incentives). Increased incentives assumed a 0.88 percent per year
chance of using insulin pumps as compared to 0.44 percent per year for the continued incentives scenario.
Increased utilization of insulin pumps contributes to a reduction in overall diabetes mortality because
a larger proportion of diabetic patients are using the pump, which reduces the diabetes death rate by
preventing complications and hypoglycemic events.
Decreased incentives
PRC for the “decreased incentives” scenario was determined through reducing the annual probability of
initiating insulin pump use to 0.22 percent. This leads to a reduction in the number of diabetes patients
using insulin pumps and contributes to an overall increase in the mortality risk for diabetes patients
because more people are using insulin injections compared to the continued incentives scenario.
Heart disease
A similar Markov model as above was specified for heart disease. The aggregate effects of these technologies
are assessed through calculation of PRC for diagnosed heart disease. After determination of the PRC,
the effects of technology on heart disease expenditures are projected.
The heart disease model had eight health states:“Well, under 35”;“Well, over 35”;“Undiagnosed heart
disease”;“Diagnosed heart disease”;“Heart disease, post-acute coronary event”;“Heart disease, post-
surgery”;“Heart disease, post-surgery and acute coronary event”; and“Dead.”“Dead”state includes dying
of heart disease and dying of other causes. Initial health state populations are based on 2010 PRC estimates
in each of the states calculated from MEPS and the CDC. Since heart disease incidence is low for age group
“under 35,”we assumed age 35 as a cutoff point. Therefore, individuals in the“Well, under 35”health state
can either die of other causes (based on CDC general mortality tables), remain in the“Well, under 35”health
state, or turn 35 (according to the U.S. Census projections). People in the“Well, over 35”state can survive or
die of other causes.
Each year, someone in the “Well, over 35” health state can obtain heart disease. The Framingham risk score
formulas were developed based on the seminal cohort studies to predict the likelihood of developing heart
disease using various risk factors for disease.49
Risk factors include age, obesity, gender, smoking status,
blood pressure, and cholesterol. All risk factors are assumed to be independent, following methods of other
49. Dagostino et al., “Primary and subsequent coronary risk appraisal: New results from the Framingham Study.” American Heart Journal 139:2.1 (2000).

79Methodology
published Markov models. The values input into the Framingham risk score formulas were average data on
risk factors obtained from the CDC data.50
Risk factors other than age and obesity are assumed to remain
constant over time. The influence of other risk factors or variations in the trajectory of risk factors is assessed
in sensitivity analysis in the incidence rate. We assume risk factors influence only the incidence of disease
because once a patient has heart disease, he or she is at an increased level of illness severity.
Heart disease can present with symptoms, but many people develop heart disease unknowingly and
without symptoms. The symptoms can either be mild, such as a stable angina, or more acute, such as
unstable angina, myocardial infarction, or cardiac arrest. Those with undiagnosed heart disease can be
diagnosed through diagnostic screening or may develop symptoms. It is assumed that all patients with
symptoms are diagnosed with disease. If disease is not diagnosed, patients go on with heart disease
but without treatment, putting them at greater risk for an acute coronary event. As such, there is an
undiagnosed heart disease health state within the model.The probability of screening is based on the
percentage of heart disease PRC with a diagnostic test but without a surgical procedure; this value was
used to approximate the number of people using heart disease technology for diagnostic purposes.
A certain percentage of people who develop heart disease are detected through identification of symptoms;
the proportion of silent to diagnosed heart disease is used as the probability of being detected by symptoms.
The probability of symptoms of angina versus acute coronary event is obtained from models based on the
Coronary Heart Disease Policy Model.51
Patients are placed into the most severe category of their illness;
therefore, if a patient has both angina and an acute coronary event, they will be placed in the post-event
category. Individuals experiencing an acute coronary event may die from the event, may require surgery,
and may die from that surgery (complications). Depending on these outcomes, patients would transition
to appropriate health states within the model. It is assumed that angina alone poses no risk of death, based
on similar assumptions from previous models. The model assumes that patients identified with the disease
obtain treatment and do not discontinue treatment.
People with heart disease may have preventive surgery to improve outcomes based on MEPS annual surgery
rates. If patients survive surgery, they would be placed in the post-surgery health state, where they would
have reduced risk for an acute event. Patients in this category can still have another event; if they survive,
they would be placed in the post-surgery and event health state.
Below is a diagram describing the potential transitions between health states. All states can lead to the
death state, and therefore it was not included in the model.
50. Data sources included National Health Interview Survey, National Health and Nutrition Examination Survey, and Behavioral Risk Factor
Surveillance System.
51. Weinstein et al., “Forecasting Coronary Heart Disease Incidence, Mortality, and Cost: The Coronary Heart Disease Policy Model.” American
Journal of Public Health 77:11 (1987).

80 Healthy Savings
Figure A3
State transition diagram for heart disease Markov model
Well,
under 35
Well,
35 and over
Post-event
and surgery
Post-acute
coronary event
Post-surgery
Heart disease,
undiagnosed
Heart disease,
diagnosed
The likelihood to obtain planned surgery, the risk of death from surgery, the likelihood to obtain a
diagnostic test, the diagnostic test sensitivity, the relative risk of a coronary event given heart disease
treatment, and the relative risk of death with diagnosed heart disease were the variables changed between
each incentive scenario. The continued incentives scenario assumes an increase in the rate of diagnostic
testing concurrent with the 2005-2010 historic rates from MEPS (an annual chance of 39 percent in 2010
and assumed to increase to 42 percent in 2035. Similar methods are used for surgery, beginning with
2.9 percent in 2010 and increasing to 4.5 percent in 2035).
The increased incentives scenario assumes increased innovation of medical technology and, with that,
improvements in the effectiveness of these technologies. Therefore the diagnostic test sensitivity was
improved and the surgical mortality risk was decreased in this scenario. Additionally, increased technology
adoption was assumed, so the likelihood to obtain a diagnostic test or surgery was increased. Because
proper diagnosis can aid treatment and help identify less severe patients as having heart disease,
we assume that the treatment is more effective at preventing acute events and that the death rate of
diagnosed heart disease decreases.
The increased incentives scenario begins with the same annual probabilities to obtain diagnostic testing
and surgical procedures. However, the rate of increase for the probability of obtaining surgery doubles. The

81Methodology
rate of increase is five times higher for the probability of obtaining a diagnostic test. This leads to a 5.2 percent
annual surgery probability and a 51 percent chance of obtaining a diagnostic test. Additionally, the relative
risk of obtaining an event with treatment is decreased by 25 percent, the relative risk of death with diagnosed
heart disease is decreased by 15 percent, probability of dying from surgery is decreased by 50 percent, and the
diagnostic test sensitivity is increased by 50 percent.
The decreased incentives scenario assumes decreased technological adoption and therefore decreased rates
of both diagnostic testing and surgery. This scenario halves the rate of increase of the probability of surgery
and reduces the rate of increase for the probability of obtaining a diagnostic test by 80 percent. This leads to
a 3.2 percent chance of surgery and 40 percent chance of diagnostic testing by 2035. Because of increasing
severity of disease, the relative risk of death with diagnosed heart disease was increased by 50 percent in
comparison to the continued incentives scenario.
The Markov model created for musculoskeletal disease assessed the effects of diagnostic MRIs and
joint replacement surgery on PRC from 2010 to 2035. The overall PRC for musculoskeletal disease was
assessed by summing the PRC for people with mild and severe disease, and projected expenditures of
musculoskeletal disease were subsequently calculated.
It was assumed that individuals can develop musculoskeletal disease only at or after age 40 since the
prevalence of the disease for the below 40 age group was minimal. The musculoskeletal disease (MSD)
model has nine health states: “Well, under 40”; “Well”; “Mild MSD, improper treatment”; “Mild MSD,
treatment”; “Severe MSD, treatment”; “MSD, post-surgery”; “MSD, post-revision”; “MSD, treatment failure”;
and “Dead,” as seen in Figure A4. Initial health state probabilities are estimated from 2010 prevalence of
these states. Musculoskeletal disease encompasses a broad range of conditions, and characteristics of
rheumatoid arthritis and osteoarthritis were combined to represent incidence and disease progression of
the larger category. Probabilities were derived from the weighted mean of the relevant variables for the
two diseases.
Individuals in the “Well, under 40” health state can survive or die (of all causes) and if they survive, they can
turn 40 and transition into the “Well” health state. We incorporated the effects of obesity and aging as risk
factors for disease into our model. Individuals who do not develop musculoskeletal disease remain in the
“Well” state.
Individuals can then obtain an MRI based on MRI utilization calculated from MEPS data. Based on the
MRI sensitivity for musculoskeletal disease, diagnostic tests can determine whether disease exists or the
extent of progression of the disease. We assume that accurate diagnostic testing allows patients with
musculoskeletal disease to be identified and given the proper treatment, whether that be a medical
treatment to prevent the progression of rheumatoid arthritis or a lifestyle modification to prevent the
progression of osteoarthritis. If an MRI was not performed, there was a chance that the disease was
symptom diagnosed and appropriate treatment was recommended. Within this model, diagnosis of disease
is assumed to be concurrent with proper treatment. Missed diagnoses and improper treatment of disease
(that will lead to failure rates) would be categorized as “Mild MSD, improper treatment.” Depending on
whether disease is identified, individuals can jump to the “Mild MSD, improper treatment” or “Mild MSD,

82 Healthy Savings
treatment” health state. MRI can help with detecting undiagnosed disease. Undiagnosed disease can occur
because patients either do not notice joint inflammation or do not recognize its severity. Therefore, MRI can
ensure proper treatment, as physicians can prescribe appropriate medical regimens and lifestyle routines
for the type and severity of disease to improve quality of life and prevent progression.
The “Mild MSD, improper treatment” health state assumes the disease is either undiagnosed or improperly
treated. An improper treatment of the disease (including an absence of treatment) can result in further
progression of the disease into a more severe stage. Individuals in this health state can die or they can
maintain mild musculoskeletal disease or progress to severe disease. Progression probability to severe
disease is based on the average duration of the mild disease stage and the percentage of people that
progress to the severe disease stage. The model assumes that severe disease is properly identified and
treated. Individuals who maintain mild disease without treatment have a chance of obtaining an MRI and
having their disease detected, or having symptoms appear that would facilitate proper treatment of the
disease. Both scenarios lead to patients moving into the “Mild MSD, treatment” health state. Otherwise,
patients remain in the “Mild MSD, improper treatment” health state.
Patients in the “Mild MSD, treatment” health state can survive or die based on CDC general mortality
tables. If patients survive, they can progress to the “Severe MSD” health state or remain in the “Mild MSD,
treatment” health state. Individuals in the “Severe MSD” health state can survive or die based on CDC
mortality rates multiplied by a relative risk of death due to musculoskeletal disease. Everyone in this health
state is assumed to have diagnosed disease and to be obtaining treatment. People with diagnosed disease
can obtain surgery based on MEPS derived surgical rates. The proportion of people with musculoskeletal
disease with a surgery-related expenditure in a given year was assumed to be the surgery rate. This was
projected out based on trends seen in 2005-2010 MEPS data. Surgery is the other technology assessed
as part of this study. If patients survive surgery, they jump to the “MSD, post-surgery” health state, and if
patients do not obtain surgery, they remain in the “Severe MSD” health state.
People in the“MSD, post-surgery”group may need a revision joint replacement surgery, which they may
survive or die from. Only one revision is assumed and the potential for revision is based on revision rates
found in scientific literature. If no revision is performed, treatment success is assumed and patients remain in
the“MSD, post-surgery”health state. According to revision success rates, patients may experience treatment
success and maintain in the“MSD, post-revision”health state or they may experience treatment failure and
jump to the“Treatment failure”health state. People in the“Treatment failure”state may survive or die.
Below is a diagram visualizing the potential transitions between health states in the model described
above. All states can transition to the “Dead” state.

83Methodology
Figure A4
State transition diagram for musculoskeletal disease Markov model
Well,
under 40
Well,
40 and over
Severe disease
Post-revision
Treatment
failure
Post-surgery
Mild disease,
improper treatment
Mild disease,
treatment
The Markov models were used to determine PRC for the disease state. To calculate expenditures,
expenditures per PRC determined from the 2005-2010 MEPS data were multiplied by the PRC and adjusted
to match the 2010 expenditures calculated from MEPS.
Three incentive scenarios are addressed in this analysis: continued incentives, increased incentives, and
decreased incentives. Variables changing between the incentives scenarios include likelihood of obtaining
a diagnostic test, the likelihood of obtaining surgery, and the relative risk of progression between mild and
severe disease with treatment.
Probability of obtaining a diagnostic test and obtaining joint replacement surgery were increased linearly
based on the trend observed in 2005-2010 MEPS data for the continued incentives scenario. Relative risk of
progression was kept at its base value for this scenario.
We assumed twice the original growth rate for diagnostic testing and undergoing surgery. We also assumed
an increase in MRI sensitivity for arthritis from 0.8 to 0.95 and a 16 percent reduction in risk of death due
to musculoskeletal disease compared to the continued incentives scenario. These adjustments are a result
of improved diagnostic technology, which improves physicians’ability to treat and improved therapeutic
technology increasing treatment effectiveness. Similarly, a 50 percent decrease in the likelihood of surgery
revision, surgery failure, and risk of death due to surgery were assumed.

84 Healthy Savings
The probabilities for obtaining an MRI or surgery were assumed to increase at half the original rate. We also
assume a 25 percent increase in relative risk of death due to musculoskeletal disease compared to the
continued incentives scenario due to more severe disease and lack of treatment options. We did not
assume a decrease in the sensitivity of current diagnostic technology or an increase in negative outcomes
from surgery because technology will not worsen in the future compared to 2010.
Colorectal cancer
Detection and treatment
A colorectal cancer model was determined to assess the effects of increased screening adoption in PRC
for colorectal cancer, as well as to determine the number of people prevented from developing colorectal
cancer due to screening.
The colorectal cancer disease model has multiple health states: “Well, under 50”; “Well, needs screen”; “Well,
post-screen”; “Post-polypectomy”; “Missed adenoma”; “Stage 1 cancer, symptom detected”; “Stage 1 cancer,
screen detected”; “Stage 2 cancer, symptom detected”; “Stage 2 cancer, screen detected”; “Stage 3 cancer,
symptom detected”; “Stage 3 cancer, screen detected”; “Stage 4 cancer, symptom detected”; “Stage 4 cancer,
screen detected”; and “Dead.” People can survive or die based on CDC general mortality rates. People with
colorectal cancer have an increased risk of death corresponding to the stage of the disease.
Since current colorectal cancer screening guidelines recommend screening for those over age 50, people under
50 are placed in a different health state and can develop cancer based on SEER age-specific colorectal cancer
incidence levels. People above the age of 50 are placed into different states based on whether they are well,
have a polyp, have had a polypectomy, or have cancer. People in the well population can develop polyps based
on the incidence rates found in scientific literature. As colorectal cancer risk increases with age and obesity
rates, the model incorporates these risk factors to the incidence of cancer and polyps.
A certain portion of the eligible population obtains screening based on current screening compliance rates.
Colonoscopies may detect polyps in patients based on the specificity of colonoscopy, and when a polyp
is detected it is removed through a polypectomy. Though colonoscopy and polypectomy both render a
risk of death, these risks are not considered in this model. Additionally, colonoscopies may also detect
cancer in patients without cancer (known as a false positive). Screening and false positives result in cost
and productivity losses, but these effects are not considered in the Markov model. (Screening expenditures
on the healthy population are calculated separately.) A history of polyps increases risk of future polyp
development. Additionally, patients who have undergone polypectomy are required to obtain surveillance
screening every three years until they obtain a negative colonoscopy, after which they can transfer to the
“Well, post-screen”state.
Patients who undergo screening with no detected polyp are placed in a post-screening state, where they
are eligible for future screening according to guidelines. They may develop a polyp during this surveillance
time, which may be detected if they obtain their future recommended screenings.
Missed polyps can progress to cancers over time, though only a portion of polyps are precancerous.
Scientists have not yet determined the average dwell times for precancerous polyps before they transition
into cancer, and the time varies between patients based on a number of risk factors.

85Methodology
The effect of screening on detecting cancer is modeled by an increased likelihood to identify cancer in
an earlier stage than it was detected through symptoms. Earlier stage cancer is easier to treat and poses a
reduced level of mortality. The likelihood distribution is obtained from scientific literature.
Once people develop cancer, they can either maintain that stage of cancer, progress to a more severe stage
of cancer, or die based on mortality risk associated with their current stage of cancer.
The diagram below describes the transitions between the potential health states incorporated into the
colorectal cancer Markov model. Both screen-detected and symptom-detected cancer states are separated
into the stages of cancer, and allowing for transition to more progressed stages. Each health state allows for
the transition into the “dead” state.
Figure A5
State transition diagram for colorectal cancer Markov model
Well,
under 50
Well,
50 and over
Well,
post-screening
Post-
polypectomy
Missed
adenoma
Screen-detected
cancer
Symptom-
detected cancer
The continued incentives scenario assumes a continuation in the current annual change in screening rates
derived from MEPS 2005-2010, beginning with a 4.2 percent chance of obtaining screening per year and
increasing to a 8.0 percent chance of screening in 2035 (screening is recommended every 10 years).
The increased incentives assumed twice the original growth rate in screening, rising to 11.8 percent of the
well population screened per year.
The decreased incentives scenario assumed one half of the original growth rate, leading to only 6.1 percent
of the screening eligible population obtaining it that year.

86 Healthy Savings
Prevention
The Markov model was also used to determine the number of cases prevented by screening. For historical
calculations, we assumed one-third of polyps would have developed into cancer, and therefore one-third
of people receiving polypectomies were saved by screening. A tracker variable was used to count the
number of people receiving polypectomies each year, and the annual number of polypectomies was
used to determine the total number of cases prevented each year. Expenditures saved by screening
were calculated by multiplying the expenditure per PRC for colorectal cancer with the number of cases
prevented each year. This process remained the same for each incentives scenario.
Projection of treatment and prevention expenditures
Aggregate expenditures for each disease were obtained by multiplying appropriate PRC by expenditure
per PRC from 2005-2010 MEPS data for all diseases except colorectal cancer. Due to a small sample size
associated with the colorectal cancer population, the historic PRC for that group is an outlier. For the
purposes of the projection, the average expenditure/PRC between 2008 and 2010 was calculated and
used for the initial expenditure calculated.
In the increased incentives scenario, an annual percentage reduction was applied to the expenditures per
PRC for technology users because improved technology was assumed to reduce use of expensive sites
of service, therefore reducing the overall cost of care. Contrastingly, an annual percentage increase was
applied to the expenditures per PRC for the decreased incentives scenario. As the prevalence of risk factors
rises in the country, more people will obtain chronic disease and those with chronic disease will likely have
more severe disease. Severe disease will be more expensive to treat, and expenditures will continue to rise if
new technology is not developed.
Table A4
Reduction in expenditure/PRC in the increased incentives scenario
Associated with technology use
DISEASE COST REDUCTION (PERCENT)
Heart disease 1.0
Musculoskeletal disease 0.6
Diabetes 0.5
Colorectal cancer -
The percentage reduction in expenditure/PRC for technology usage for each disease was assigned through
an ordinal comparison. Heart disease was assumed to have the highest potential for cost reduction.This was
assumed because of the opportunity for a shift toward cheaper diagnostics, less expensive minimally invasive
techniques, and the potential for new technological development in relation to stents, pacemakers, and more.
Musculoskeletal disease is assumed to have a smaller reduction than heart disease but a greater reduction
than diabetes. One of the technologies assessed is joint replacement surgery, which can become less invasive
and have technological improvements in materials. However, minimally invasive joint replacement surgery
is less pervasive of a technique. We assumed diabetes would have the next smallest potential for reduction

87Methodology
because the only technology assessed was the pump, which serves only the insulin dependent diabetes
patients. Colorectal cancer was associated with no cost reduction associated with technology usage because
of the highly variable cost of cancer treatment.
Projection of indirect impact (foregone GDP)
In this part of the study, we extend our findings from the previous section to project future indirect impact.
Indirect impact is projected through 2035 under three alternative scenarios—the continued incentives,
the increased incentives, and the decreased incentives.
In developing the alternative scenarios of future indirect impact, we first project the future path of
employed population reporting a condition (EPRC) and employed caregivers by condition (ECC) using
employment projections from Economy.com, the U.S. Census, and the population reporting a condition
(PRC) calculated in the Markov models.
Next, we use employment and population projections to calculate employment-to-population (E/P) ratios.
Total population is calculated as those age 16 and older. Next, the E/P ratio for every year is divided by
that for 2010 to build an E/P index. For example, the E/P index for 2011 was derived by dividing the 2011
employment-to-population (0.534) by the 2010 ratio (0.533).
We then create a PRC index for each disease under each scenario. This is done by dividing PRC for every
year by the PRC for 2010. The E/P index is then applied to the PRC index to create a new “E/P-PRC index.”
This index is scaled to the 2010 EPRC and ECC to obtain projections through 2035. Lost workdays were
scaled to the 2010 ratio of lost workdays to EPRC and applied to the current year EPRC. Absenteeism and
presenteeism were then calculated in a manner similar to the methodology used to estimate the historical
indirect impact.
As mentioned above, PRC is used for each disease to calculate a new EPRC. Absenteeism and presenteeism
estimates are consistent with the methodology used for historical indirect impact.
Similar to the continued incentives, we use the PRC for each disease to calculate the EPRC. However, we use
the PRC calculated in the increased incentives scenario projections. Absenteeism and presenteeism are then
estimated via methods consistent with the continued incentives technology. Further adjustments are made
under this scenario to account for advancements in new technology, as highlighted in Table A5 below.
We adjusted the absenteeism and presenteeism loss for patients who used medical technology in the
increased incentives scenario because we assumed that this technology would improve quality of life,
therefore increasing employed patients’ ability to work, and ultimately decreasing loss to the GDP.
Because no data existed to inform a future improvement in productivity due to technology, we assumed
an ordinal increase based on disease biology.
Colorectal cancer treatment costs are highly variable, and new technology that is effective in treating cancer
may still take a large toll on patient and caregiver productivity. We therefore assumed that all reductions in
indirect impact due to increased technology development were associated with the reduction in PRC.

88 Healthy Savings
Of all diseases, musculoskeletal disease has the greatest quality of life impact for patients and caregivers,
and consequently the greatest opportunity for improvement in lost workdays. Therefore, musculoskeletal
disease lost workdays for the individual were decreased the most. When looking at historical data, insulin
pump has the smallest savings in indirect impact per affected person of all the examined diseases.
Therefore we assumed the smallest reduction in lost workdays after increased technology development.
Heart disease savings were historically in between diabetes and musculoskeletal disease, but closer to
diabetes. The technologies for heart disease are much more diverse than those for diabetes; therefore,
we assume a greater opportunity for reduction in lost workdays.
Another aspect is presenteeism, or reduced productivity of employees at the workplace. Presenteeism is
often affected by disease-related stress and apprehension. Because a consequence of heart disease is often
sudden, potentially fatal effects, presenteeism associated with heart disease is high. Improved treatment
would prevent acute coronary events, reducing fear and associated productivity loss the most of all diseases
examined. Improper management of diabetes also has the potential to cause serious hypoglycemic events,
while musculoskeletal disease is rarely the cause of an acute fatal occurrence. As such, improvements
in musculoskeletal disease technology were assigned the lowest improvement in productivity loss with
diabetes between heart and musculoskeletal diseases.
Caregivers are also significantly affected by improvements in disease-related technology; they often have
to miss work to take care of patients, the stress of which can contribute to presenteeism as well. Caregiver
lost workdays were assumed to be reduced in the same ordinal fashion to individual lost workdays after
improvements in technology in the increased incentives scenario. Heart disease caregiver presenteeism
was also assumed to be the most reduced with technology improvement. However, we assumed that
presenteeism of caregivers for musculoskeletal disease patients would decrease more than that for diabetes
simply because musculoskeletal disease’s greater indirect impact provides a higher potential for reduction.
Table A5
Adjustments for varied scenarios
Relative to the continued incentives scenario
INSULIN
PUMPS
HEART DISEASE
TECHNOLOGY
MUSCULOSKELETAL
DISEASE TECHNOLOGY
COLORECTAL
CANCER SAVINGS
Increased
incentives
Individual lost workdays 0.95 0.99 0.90 1.00
Individual presenteeism 0.87 0.947 0.95 1.00
Caregiver lost workdays 0.92 0.95 0.87 1.00
Caregiver presenteeism 0.85 0.92 0.90 1.00
Decreased
incentives
Individual lost workdays 1.01 1.02 1.03 1.00
Individual presenteeism 1.03 1.04 1.02 1.00
Caregiver lost workdays 1.005 1.01 1.02 1.00
Caregiver presenteeism 1.02 1.03 1.01 1.00

89Methodology
Under the decreased incentives scenario, PRC for decreased incentives is used to calculate a new EPRC.
Absenteeism and presenteeism are then estimated similar to the methodology for historical indirect
impacts. Similar to the increased incentives scenario, the decreased incentives assume a percentage
increase in presenteeism and lost workdays as seen in the above table. Worsening risk factors increase the
severity of the disease and a lack of incentives to innovate new technology would exacerbate the problem.
Survival
While in some cases increased adoption of beneficial medical technology can decrease the economic
burden associated with a disease, our projections reveal that it can also increase the economic burden of
a disease. This increase in economic impact of disease is related to increased PRC due to increased survival
from better medical treatment. Therefore, changes in survival associated with differing incentives scenarios
were calculated to give further insight into the drivers of the changes in economic impact.
To determine changes in the survival, we first take the cumulative sum of those who have died thus far in
the model by calculating the PRC of the “Dead” health state for each projected year. Then the differences
in PRCs between the continued incentives and increased incentives scenarios and between the continued
incentives and the decreased incentives scenarios were calculated. These differences represent the
extra population that is dead or alive due to the changes in incentives scenarios. We then calculated the
proportional change in PRC due to changes in survival. The continued-increased difference in PRC who
have died was divided by the disease PRC for the increased incentives scenario for each projected year.
This gave the proportion of the PRC who was alive in the increased incentives scenario compared to the
continued incentives scenario. The continued-decreased difference in PRC who have died was divided by
the disease PRC in the decreased incentives scenario for each projected year. This gave the ratio of extra
deaths due to the decreased incentives scenario to the disease PRC in the decreased incentives scenario.
These ratios were then multiplied by the total indirect impact of the disease to determine the amount
impact survival has on increases or decreases in productivity.
Healthy people screened
For the primary analysis for this report, we calculated the economic burden associated with disease in three
scenarios with differing rates of medical technology adoption. Some of the technologies, such as EKG for
heart disease, colonoscopy, and MRI for musculoskeletal disease, are for screening. Increased screening
rates associated with increased adoption of these technologies entails increased expenditures associated
with the healthy population as well. We wanted to capture these costs in order to present a balanced
analysis of the effects of increased technology adoption.
Heart disease
Historical calculations
First, the unique number of people receiving EKGs was calculated from MEPS in 2010. The PRC for people
with heart disease receiving EKGs was then subtracted from this value, yielding the number of healthy
people who received an EKG without a heart disease diagnosis. We considered this the number of healthy

90 Healthy Savings
people screened by EKG in 2010. Then the ratio of healthy people given EKGs to people with heart disease
given EKGs was calculated by dividing the two numbers.This ratio was then multiplied by the PRC for those with
heart disease receiving sonograms and the PRC for those with heart disease receiving X-rays to determine the
number of healthy people screened using sonograms and X-rays as diagnostic tests. The ratio calculated from
EKG utilization was applied to the other diagnostic tests because EKG is most closely related to heart disease.
The number of healthy people screened with each diagnostic test was multiplied by the cost of the diagnostic
test to calculate the cost of screening. There can be significant geographic variation in the expenditures
associated with individual diagnostic tests, and therefore these costs could be subject to change.
Projections
We also wanted to project the number of healthy people screened in the future. To do this, we first obtained
the number of people in the healthy population from the models, using the combination of the“Well, 35+”
and the“Heart disease, undiagnosed”health states. The 2010 rate of screening for the healthy population
was calculated by dividing the number of healthy people screened calculated from historical data by the
size of the healthy population. The rate was projected to increase proportionally to the rate increase in the
projection model for each scenario. These rates were multiplied by the projected healthy population size
to determine the number of healthy people screened in the future. The future cost of screening the healthy
population was calculated by multiplying the average expenditures associated with screening by the
number of healthy people screened. Expenditures were then multiplied by the consumer price index (CPI)
to account for inflation.
First, we calculated the total number of people receiving an MRI in 2010 from MEPS. The PRC for people
receiving an MRI and diagnosed with musculoskeletal disease was then subtracted from the total number of
people receiving an MRI to yield the number of healthy people screened using an MRI. Then we calculated
the ratio of healthy people receiving an MRI to people with musculoskeletal disease receiving an MRI by
dividing the two numbers. This ratio was multiplied by the PRC for those with musculoskeletal disease
receiving an MRI in the previous years to calculate the historic number of healthy people screened.The cost
of screening the healthy population was calculated through multiplying the average cost of MRI by the
number of healthy people screened.
Projections
The size of the healthy population in future years was obtained from the model. The proportion screened
for the healthy population in 2010 was calculated by dividing the PRC of healthy people receiving an
MRI calculated from the historical data by the PRC for the total healthy population found in the model.
The proportion of healthy people screened was then increased at the same rate of increase seen for
the proportion of those with musculoskeletal disease screened. This proportion was multiplied by the
projected PRC for the healthy population to obtain the number of healthy people screened. The cost of
screening the healthy population was obtained by multiplying the number of healthy people screened by
the average cost of MRI. Aggregate expenditures were then multiplied by the CPI to account for inflation.

91Methodology
Colorectal cancer
Using MEPS, we calculated the number of people receiving a colonoscopy in 2010; this number was used to
approximate the number of colonoscopies per year. From the number of total colonoscopies, the number
of people with colorectal cancer and a colonoscopy and the number of cases prevented were subtracted
to obtain the number of healthy people screened for colorectal cancer in 2010. This number was then
divided by the PRC for colonoscopy and colorectal cancer in 2010, to obtain the ratio of healthy screening
colonoscopies to colorectal cancer colonoscopies. This ratio was then applied to the historical years
assessed to obtain the historical numbers for healthy people screened. To obtain the expenditures associated
with screening the healthy population, the average cost of a colonoscopy was multiplied by the size of the
healthy population screened.
Projections
The projected size of the healthy population was obtained.The proportion screened for the healthy population
in 2010 was calculated by dividing the number of healthy people screened by the size of the total healthy
population.The proportion of healthy people screened was then increased at the rate used within the model
for colonoscopy. Then the proportion for each projected year was multiplied by the projected size of the
healthy population to obtain the number of healthy people screened in the future. The cost of screening
the healthy population was obtained by multiplying the number of healthy people screened by the average
colonoscopy cost, and inflation was accounted for by multiplying these values by the projected CPI.
Validation
Table A6
Comparison of final PRC projections with outside sources
(millions)
DISEASE
MILKEN INSTITUTE
STUDY
OTHER
STUDIES
SOURCE OF
OTHER STUDIES
Diabetes 55.59 54.1-63.6 Centers for Disease Control and Prevention
Heart disease 38.87 39.59 American Heart Association
Musculoskeletal disease* 66.59 66.96 Centers for Disease Control and Prevention
Colorectal cancer** 1.69 1.5 National Cancer Institute
* CDC estimate is for arthritis alone.
** PRCs used for validation are from 2020 not 2035.

92 Healthy Savings
After the final PRC were calculated from the Markov models, we compared them to other published
numbers to ensure their similarity. This helped us to validate our results. As seen in the above table,
the PRCs generated by the model closely match those published in peer-reviewed journals by other
expert organizations for each disease.
Sensitivity analysis
Sensitivity analyses were performed on all the models in order to determine reasonable intervals for PRC
projections. Since not all data informing model creation exists in the literature, some numerical inputs
were based on approximations or assumptions. Reasonable upper and lower bounds were estimated
for variables based on assumption or subject to change. These bounds were entered into the model to
determine maximum and minimum PRCs. Sensitivity analysis was performed for each scenario to assess
how increasing or decreasing incentives would affect PRC.
Table A7
2035 Diabetes PRC based on sensitivity analysis
(millions)
CONTINUED
INCENTIVES
INCREASED
INCENTIVES
DECREASED
INCENTIVES
Base case 55.59 55.65 55.57
100% increase in incidence rate 88.80 88.88 88.75
50% decrease in incidence rate 36.14 36.18 36.13
100% increase in relative risk of mortality 44.19 44.25 44.16
50% decrease in relative risk of mortality 62.88 62.92 62.86
Two primary aspects of the diabetes Markov model were analyzed for sensitivity. Diabetes incidence is
projected to increase in the future due to increasing incidence of many of the key risk factors including
aging, obesity, and high cholesterol. However, it is difficult to predict how these risk factors will change and
changes in these risk factors will most likely be related to each other; therefore, an analysis was performed
to obtain bounds for the data if the incidence doubled or decreased by half. Changing incidence affects
overall diabetes PRC and total expenditures, as well as PRC and expenditures associated with insulin pumps.
Relative mortality risk associated with diabetes was also investigated because multiple estimates have been
published for this variable.
Doubling the diabetes incidence produces the most drastic effects in PRC between scenarios, as an
increased number of people would be insulin-dependent and an increased proportion of those people
would be helped by insulin pump usage. Regardless of changing variables, the trends between incentives
scenarios remain the same.

93Methodology
Table A8
2035 Heart disease PRC based on sensitivity analysis
(millions)
CONTINUED
INCENTIVES
INCREASED
INCENTIVES
DECREASED
INCENTIVES
Base case 38.87 41.79 38.28
Heart disease incidence was assessed to check for the effects of potential changes in the levels of risk factors;
for example, obesity in the United States is increasing while the rate of smoking is on the decline. Mortality
risk is also considered due to increased risk factors increasing severity of disease. The most drastic effect
involved changing the incidence rate, and all trends between incentives scenarios remained the same.
Table A9
2035 Musculoskeletal disease PRC based on sensitivity analysis
(millions)
CONTINUED
INCENTIVES
INCREASED
INCENTIVES
DECREASED
INCENTIVES
Base case 66.59 67.32 65.68
The primary variables assessed for sensitivity within the musculoskeletal disease model included the incidence
rate of developing musculoskeletal disease and the probability of disease progression. This analyzes how the
PRC would be affected by varying levels of risk factors to developing musculoskeletal disease, which would
increase or decrease the incidence rate and/or speed of progression. Results of the sensitivity analysis showed
that expected trends between the incentives scenarios did not change when the variables were altered.

94 Healthy Savings
Table A10
2035 Colorectal cancer PRC based on sensitivity analysis
(millions)
CONTINUED
INCENTIVES
INCREASED
INCENTIVES
DECREASED
INCENTIVES
Base case 1.69 1.41 1.85
100% increase in incidence 2.56 2.14 2.81
50% decrease in incidence 1.20 0.88 1.17
100% increase in adenoma transition rate 2.97 2.48 3.27
50% decrease in adenoma transition rate 0.94 0.80 1.03
Risk of developing colorectal cancer was analyzed at varying levels to assess how changes in risk factors
would affect projected PRCs. Additionally, the rate of progression from adenoma to colorectal cancer was
assessed due to the uncertainty surrounding this variable in the literature. Changing these inputs altered
PRCs in an expected manner, with the increased incentives scenario always yielding a lower PRC compared
to the continued incentives scenario and the decreased incentives scenario yielding a higher PRC.

95Methodology
Variable Input Tables
Table A11
Diabetes Markov model variable inputs
NAME VALUE NOTE SOURCE
Age proportions U.S. Census
Relative risk of death from non-insulin
dependent diabetes
1.688 Specific value is assumption
Boyle et al., 2010; Schuffham and
Carr, 2003
Relative risk of death from diabetes with
insulin injections
Carr, 2004
Relative risk of death from diabetes with
insulin pump
Carr, 2005
General mortality 0.0078
Centers for Disease Control and
Prevention
Diabetes incidence rate* 0.0069 Age-dependent
Prevention
Probability of obesity* 0.287
Future trends projected
from historic data
Prevention
Probability of high cholesterol* 0.384
Future trends projected
from historic data
Prevention
Probability of needing insulin at incidence 0.110
Assumption based on
proportion of insulin-
dependent diabetics
Meigs et al., 2003; Ramlo-Halsted
et al., 2000; Gess et al., 2006
Probability of insulin pump if new insulin user 0.070 Assumed to be constant
Bode et al., Medtronic
presentation
Annual probability of beginning insulin pump
usage if using insulin injections*† 0.0044
Calculated from multiple
sources
Bode et al., 2002; MEPS historic
data, Centers for Disease Control
and Prevention
Annual probability of becoming insulin-
dependent
0.012 Medical Expenditure Panel Survey
Relative risk of diabetes for obese 1.857
sources
Chan, Haffner et al., 1990; Wilson
et al., 2007; Bang et al., 2009
Relative risk of diabetes for high cholesterol 2.068
sources
Wilson et al., 2007
* 2010 probability.
† Variable changed in increased and decreased incentives scenarios.

96 Healthy Savings
Table A12
Heart disease Markov model variable inputs
Ageproportions U.S. Census
Generalmortalityrate 0.008
Prevention
Probabilityofobesity* 0.278
Future trends based on historic
data
Prevention
Heartdiseaseincidence*
Calculated from Framingham
study prediction model, based on
2010 variable inputs from CDC and
NHANES
D'Agostino et al., 2000; Centers for
Disease Control and Prevention
Relativeriskofheartdiseasewithobesity 1.780
Probabilityofsymptom-detectionofheart
disease
0.250
Probabilitythesymptomdetectedisanacute
coronaryevent(versusangina)
0.514
Weinstein et al., 1987; Bonneux et
al., 1994
Probabilityofdiagnostictest*†
0.394
data
Diagnostictestsensitivity†
0.800 Assumption
Probabilityofdeathfromacutecoronaryevent 0.240 Bonneux et al., 1994
Probabilityofsurgeryinfirstyearfollowingacute
coronaryevent
0.026 Bonneux et al., 1994
Probabilityofdyingfromsurgery†
Relativeriskofdeathwithheartdisease†
1.100
al., 1994
Probabilityofacutecoronaryeventwithheart
disease
0.010
al., 1995
Relativeriskofacutecoronaryeventwithheart
diseasetreatment† 0.052 Gaziano et al., 2006
Probabilityofpreventivesurgery*†
0.021
data
Relativeriskofacutecoronaryeventafter
surgery
Probabilityofrecurrentacutecoronaryevent 0.071 Bonneux et al., 1994
Initialproportionofundiagnosedheartdiseaseof
allheartdisease
0.229 Airaksinen and Koistinen, 1992
Initialproportionofallpeoplethatarepost-event 0.025
Myocardial infarction is assumed
to be main contributer to cardiac
events
Heart Disease Foundation
Initial proportion of heart disease patients
post-surgery
0.0019
Assume proportion of population
post-surgery is equal to MEPS
2010 data
Initial proportion of heart disease patients
post-surgery and event
0.0019
Assume equal proportion of
population has had both surgery
and event
* 2010 probability.

97Methodology
Table A13
Musculoskeletal disease Markov model variable inputs
General mortality rate 0.0078
Prevention
Future trends based on
historic data
Prevention
Incidence of musculoskeletal disease 0.015
sources
Prevention
Relative risk of musculoskeletal disease with
obesity
1.740
sources
Prevention
Probability of MRI*†
0.063
historic data
Medical Expenditure Panel
Survey
Probability of symptom diagnosis at disease
incidence
0.500
Assumption that half of
musculoskeletal disease is
symptom diagnosed
Mild to severe disease transition probability†
0.049
Based on average ten
year transition period and
assumption that half of
people transition to more
severe disease
Suter et al., 2011; Welsing et al.,
2006
Relative risk of disease transition with
treatment† 0.700
sources
Suter et al., 2011
Probability of symptom diagnosis after
disease incidence
0.035
Assumption that half of mild
disease will be diagnosed over
the course of 14 years
Suter et al., 2011; Welsing et al.,
2006
Probability of surgery*†
0.147
historic data
Survey
Probability of death from surgery†
0.0063 Brown et al. 2011; Ruiz et al., 2013
Probability of surgery failure†
0.024 Brown et al. 2011; Ruiz et al., 2013
Relative risk of mortality with
musculoskeletal disease† 1.500
sources
Prevention
Initial proportion with severe disease of total
population
0.109
Assumed equal to the number
of people with MRI or surgery
Survey
Initial proportion post-surgery of severe
patients
0.165
Surgery only patients from
MEPS
Survey
Initial proportion post-revision of surgery
patients
0.250 Assumption Brown et al. 2011; Ruiz et al., 2013
Initial proportion of post-surgery patients
with treatment failure
0.050 Assumption Brown et al. 2011; Ruiz et al., 2014
* 2010 probability.

98 Healthy Savings
Table A14
Colorectal cancer Markov model variable inputs
General mortality rate 0.008
Prevention
historic data
Prevention
Relative risk of colorectal cancer with obesity 1.110
Vogelaar et al., 2006; National Cancer
Institute
Incidence of colorectal cancer under 50 0.000092 National Cancer Institute
Relative risk of adenoma with obesity 1.180
Vogelaar et al., 2006; National Cancer
Institute
Adenoma incidence with no history 0.020 Heitman et al., 2010
Adenoma incidence with history 0.038 Heitman et al., 2010
Annual probability of screening*†
0.042
historic data
Colonoscopy specificity†
0.800
Winawer et al., 1997; Sonnenberg et
al., 2002
Screening interval after negative screen 10 years
Prevention
Screening interval after polypectomy 3 years
Prevention
Compliance probability with follow-up
polypectomy screening
0.630 Heitman et al., 2010
Dwell time 30 years
Assumptionthat1/3ofcancers
developfromadenomas
Loeve et al., 1999; Winawer et al., 1997
Initial cancer stage proportions if screen-detected
Stage 1 0.425 Heitman et al., 2010
Initial cancer stage proportions if symptom-detected
Dwell time in each cancer stage 3 years
Loeve et al., 1999; Winawer et al.,
1997; Heitman et al., 2010
5-year survival
Initial proportion of cancer that was screen
detected
0.223 Ramsey et al., 2003
Initial proportion of entire population post-
polypectomy
0.017
Healthcare Cost and Utilization
Project
Initial proportion of screening age population
with no adenoma
0.776 Heitman et al., 2010
* 2010 probability.

99Methodology
A complete Markov model diagram for diabetes
Figure A6
Incentives scenarios and health states
No diabetes
P=0.982
Diabetes, no insulin
P=0.054
P=0.016
P=0.0012
P=0
Diabetes, insulin injections
Diabetes, insulin pump
Diabetes Dead
M
M
M
Figure A7
No diabetes (part 1)
P=0.374
P=0.626
P=0.374
P=0.626
Cholesterol
not high
High
cholesterol
Cholesterol
not high
High
cholesterol
Obese*
P=0.278
Not obese
P=0.972
Survive
P=0.993
Die, other
P=0.007
No diabetes
P=0.928

100 Healthy Savings
Figure A8
No diabetes (part 2)
P=0.374P=0.374
High
cholesterol
High
cholesterol
Obese*
P=0.278
P=0.929
P=0.071
Injections
Pump
P=0.110
No insulin
P=0.890
Insulin
P=0.929
P=0.071
Injections
Pump
P=0.890
Insulin
P=0.110
No insulin
No diabetes
No diabetes
P=0.9927
Remain
diabetes free
P=0.0073
P=0.626
Cholesterol
not high
Develop
diabetes
P=0.9930
Remain
diabetes free
P=0.0070
Develop
diabetes
*Continuation from obese chance node in Figure A7.
Figure A9
Non-insulin dependent diabetes
P=0.929
P=0.071
Insulin injections
Insulin Pump
P=0.988
Remain
without
insulin
Dead
P=0.012
Progress to
insulin
P=0.013
Die
P=0.987
Survive
P=0.054
Non-insulin
dependent
diabetes

101Methodology
Figure A10
Insulin dependent diabetes using insulin injection, insulin dependent diabetes using
insulin pump, and dead
P=0.995
P=0.995
Remain on
injections
Dead
Diabetes, Insulin Pump
Dead
Transfer to
pump
P=0.984
P=0.016
Survive
Die
Diabetes, Insulin Pump
Diabetes, Insulin injections
P=0.020
Die
P=0.980
Survive
P=0.016
P=0.0012
P=0
Dead
Insulin
dependent
diabetes using
insulin injections
Insulin
dependent
diabetes using
insulin pump

102 Healthy Savings
Table A15
CCS and ICD-9 codes used for analysis
DISEASE
APPLICABLE
CCS/ICD-9 CODE
DESCRIPTION
Diabetes
49 Diabetes mellitus without complication
50 Diabetes mellitus with complication
Heart disease
96 Heart valve disorders
97
Peri-; endo-; and myocarditis; cardiomyopathy (except caused by
tuberculosis or sexually transmitted disease)
100 Acute myocardial infarction
101 Coronary atherosclerosis and other heart diseases
102 Nonspecific chest pain
103 Pulmonary heart disease
104 Other and ill-defined heart disease
105 Conduction disorders
106 Cardiac dysrhythmias
107 Cardiac arrest and ventricular fibrillation
108 Congestive heart failure; nonhypertensive
Musculoskeletal
diseases
201
Infective arthritis and osteomyelitis (except caused by tuberculosis or
sexually transmitted diseases)
202 Rheumatoid arthritis and related diseases
203 Osteoarthritis
204 Other non-traumatic joint disorders
225 Joint disorders and dislocations; trauma-related
226 Fracture of neck of femur (hip)
230 Fracture of lower limb
Colorectal cancer
14 Cancer of colon
15 Cancer of rectum and anus
ICD-9 procedures
00
Procedures and interventions not elsewhere classified; therapeutic
ultrasound, other hip and knee procedures
36
Operations on vessels of the heart; includes open chest artery
angioplasty, other heart revascularization
37
Other operations on heart and pericardium; includes insertion,
replacement, removal, and revision of pacemaker device,
echocardiography
81 Repair and plastic operations on joint structures

103
ABOUT THE AUTHORS
Anusuya Chatterjee is a senior economist and associate director of research at the Milken Institute. She has
expertise in disease prevention and wellness, longevity, and productivity and emphasizes issues related to
obesity, chronic disease, and aging in her research. She is the lead author of some of the Institute’s highest-
profile publications, including“Best Cities for Successful Aging,”“Waistlines of the World,”and“Checkup Time:
Chronic Disease and Wellness in America.”Chatterjee’s opinion articles have been published in news outlets
such as Forbes and the San Diego Union-Tribune, and she is frequently quoted as an expert in mainstream
media. Her work has been cited by the“PBS NewsHour,”the Wall Street Journal, CNN, CBS, the Huffington
Post, the Los Angeles Times, and many other outlets. Previously, Chatterjee held a tenure track academic
position. She received a Ph.D. in economics from the State University of New York at Albany, a master’s
degree from the Delhi School of Economics, and a bachelor’s from Jadavpur University in India.
Jaque King is a research analyst at the Milken Institute. She is interested in economic issues specific to aging
populations, health-care reform, the impact of funding biosciences, and public policy. Recently, she presented
at the 2014 AcademyHealth Annual Research Meeting. She also coauthored“Checkup Time: Chronic Disease
and Wellness in America,” which measures the economic impact of chronic diseases and compares it to
projections made in the Institute’s groundbreaking report“An Unhealthy America: The Economic Burden of
Chronic Disease.”She has also contributed to the publications“Best Cities for Successful Aging,”“Waistlines of
the World,”and “Estimating Long-Term Economic Returns of NIH Funding on Output in the Biosciences.”
Previously, she was a senior editor at the Pepperdine Policy Review and authored a journal article that analyzed
the politics surrounding drug policies. Her past research projects included analyzing methods for financing the
Affordable Care Act and assessing the economics of criminal-justice policy toward nonviolent drug offenders.
King holds a master’s of public policy degree with a specialization in economics and American politics from
Pepperdine University and a bachelor’s degree in political science from San Diego State University.
Sindhu Kubendran is a research/health analyst at the Milken Institute who focuses on areas of public
health that include prevention, wellness, chronic disease, and longevity. Her goal is to use data to inform
decision making and identify more effective systems of care. At the Institute, Kubendran is a co-author
of the report “Checkup Time: Chronic Disease and Wellness in America,”which compares trends in the
economic burden of chronic disease. She presented the paper at the 2014 International Health Economics
Association World Congress. Her past research includes working with a University of California, Berkeley,
research group to assess the environmental and health effects of the BP Deepwater Horizon oil spill. She
has also worked in chronic disease prevention and systems improvement at community health centers and
social service agencies. Kubendran holds a master’s of public health degree with a focus on health services
research from Dartmouth College and a bachelor’s degree in environmental engineering from UC Berkeley.
Ross DeVol is chief research officer at the Milken Institute. He oversees research on international, national,
and comparative regional growth performance; technology and its impact on regional and national
economies; access to capital and its role in economic growth and job creation; and health-related topics.
He was the principal author of “The Global Biomedical Industry: Preserving U.S. Leadership,” a study that
showed that the United States is still the global leader in the biomedical industry, but countries across
Europe and Asia are pursuing aggressive plans to close the gap and take the high-value jobs and capital this
sector creates. He was also the principal author of “An Unhealthy America: The Economic Burden of Chronic
Disease,” which brought to light the economic losses associated with preventable illnesses and estimated
the costs avoided if a serious effort were made to improve Americans’ health. DeVol is ranked among the
“Super Stars” of Think Tank Scholars by International Economy magazine. He was previously senior vice
president of IHS Global Insight.

1250 Fourth Street
Santa Monica, CA 90401
Phone: 310-570-4600
E-mail: info@milkeninstitute.org • www.milkeninstitute.org
1101 New York Avenue NW, Suite 620
Washington, DC 20005
Phone: 202-336-8930
137 Market Street #10-02
Singapore 048944
Phone: 65-9457-0212

Healthy Savings. Medical Technology and the Economic Burden of Disease

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Healthy Savings. Medical Technology and the Economic Burden of Disease (20)

More from Revital (Tali) Hirsch (8)

Recently uploaded (20)

Healthy Savings. Medical Technology and the Economic Burden of Disease