Simulating and Evaluating Local Interventions to Improve Cardiovascular Health

Homer Consulting, Voorhees, NJ 08043, USA.
Preventing chronic disease (Impact Factor: 2.12). 01/2010; 7(1):A18.
Source: PubMed
ABSTRACT
Numerous local interventions for cardiovascular disease are available, but resources to deliver them are limited. Identifying the most effective interventions is challenging because cardiovascular risks develop through causal pathways and gradual accumulations that defy simple calculation. We created a simulation model for evaluating multiple approaches to preventing and managing cardiovascular risks. The model incorporates data from many sources to represent all US adults who have never had a cardiovascular event. It simulates trajectories for the leading direct and indirect risk factors from 1990 to 2040 and evaluates 19 interventions. The main outcomes are first-time cardiovascular events and consequent deaths, as well as total consequence costs, which combine medical expenditures and productivity costs associated with cardiovascular events and risk factors. We used sensitivity analyses to examine the significance of uncertain parameters. A base case scenario shows that population turnover and aging strongly influence the future trajectories of several risk factors. At least 15 of 19 interventions are potentially cost saving and could reduce deaths from first cardiovascular events by approximately 20% and total consequence costs by 26%. Some interventions act quickly to reduce deaths, while others more gradually reduce costs related to risk factors. Although the model is still evolving, the simulated experiments reported here can inform policy and spending decisions.

Full-text

Available from: Jack Homer
VOLUME 7: NO. 1 JANUARY 2010
Simulating and Evaluating Local
Interventions to Improve
Cardiovascular Health
SPECIAL TOPIC
Suggested citation for this article: Homer J, Milstein B,
Wile K, Trogdon J, Huang P, Labarthe D, et al. Simulating
and evaluating local interventions to improve cardiovas-
cular health. Prev Chronic Dis 2010;7(1). http://www.cdc.
gov/pcd/issues/2010/jan/08_0231.htm. Accessed [date].
PEER REVIEWED
Abstract
Numerous local interventions for cardiovascular disease
are available, but resources to deliver them are limited.
Identifying the most effective interventions is challeng-
ing because cardiovascular risks develop through causal
pathways and gradual accumulations that defy simple
calculation. We created a simulation model for evaluating
multiple approaches to preventing and managing cardio-
vascular risks. The model incorporates data from many
sources to represent all US adults who have never had
a cardiovascular event. It simulates trajectories for the
leading direct and indirect risk factors from 1990 to 2040
and evaluates 19 interventions. The main outcomes are
first-time cardiovascular events and consequent deaths,
as well as total consequence costs, which combine medi-
cal expenditures and productivity costs associated with
cardiovascular events and risk factors. We used sensitivity
analyses to examine the significance of uncertain param-
eters. A base case scenario shows that population turn-
over and aging strongly influence the future trajectories
of several risk factors. At least 15 of 19 interventions are
potentially cost saving and could reduce deaths from first
cardiovascular events by approximately 20% and total
consequence costs by 26%. Some interventions act quickly
to reduce deaths, while others more gradually reduce costs
related to risk factors. Although the model is still evolving,
the simulated experiments reported here can inform policy
and spending decisions.
Introduction
Conditions in particular neighborhoods or cities can pro-
foundly enhance or impede people’s prospects for a healthy
life (1). This dependence on local context is especially evi-
dent in cardiovascular health, for which behavioral, social,
and environmental factors combine to affect the likelihood
of developing disease or dying prematurely (2). Heart dis-
ease and stroke are largely preventable, but they remain
the first and third leading causes of death in the United
States, partly because we have yet to establish living con-
ditions that minimize such modifiable risks as smoking,
obesity, stress, air pollution, poor diet, and physical inac-
tivity. The importance of intervening to limit these risks
is highlighted in A Public Health Action Plan to Prevent
Heart Disease and Stroke (3).
The notion that cardiovascular disease (CVD) can be
prevented through local actions raises practical questions
that can be examined through systems modeling and sim-
ulation. Working closely with colleagues in Austin/Travis
County, Texas, and subject matter experts at the Centers
for Disease Control and Prevention and the National
Institutes of Health, we developed a system dynamics
simulation model to answer the following questions:
• How does local context affect the major risk factors for
CVD and, in turn, population health status and costs?
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services, the
Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only and
does not imply endorsement by any of the groups named above.
www.cdc.gov/pcd/issues/2010/jan/08_0231.htm • Centers for Disease Control and Prevention 1
Jack Homer, PhD; Bobby Milstein, PhD, MPH; Kristina Wile, MS; Justin Trogdon, PhD; Philip Huang, MD, MPH;
Darwin Labarthe, MD, MPH, PhD; Diane Orenstein, PhD
Page 1
VOLUME 7: NO. 1
JANUARY 2010
2 Centers for Disease Control and Prevention • www.cdc.gov/pcd/issues/2010/jan/08_0231.htm
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services, the
Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only and
does not imply endorsement by any of the groups named above.
• How might local interventions affect CVD risk, health
status, and costs over time?
• How might local health leaders better balance their
policy efforts given limited resources?
Methods
System dynamics models improve our ability to antici-
pate the likely effects of interventions in dynamically
complex situations, where the pathways from interven-
tions to outcomes may be indirect, delayed, and possibly
affected by nonlinearities or feedback loops (4). System
dynamics has been used effectively since the 1970s to
model many areas of public health and social policy,
including CVD (5).
Model structure
We previously described a framework for understanding
cardiovascular health in a local context (6). That frame-
work has been refined and quantified by using additional
literature and input from veteran health planners and
analysts. The resulting simulation model (Figure 1) focus-
es on primary prevention; it does not address people who
have experienced a CVD event. Causal influences move
down and to the right, ending with 2 outcomes: 1) first-
time cardiovascular events and consequent deaths and 2)
Figure 1. Simulation model for cardiovascular disease (CVD) outcomes. This diagram depicts major health conditions related to CVD and their causes. Boxes
identify risk factor prevalence rates modeled as dynamic stocks. The population flows associated with these stocks — including people entering the adult popu-
lation, entering the next age category, immigration, risk factor incidence, recovery, cardiovascular event survival, and death — are not shown.
Key:
Blue solid arrows: causal linkages affecting risk factors and cardiovascular events and deaths.
Brown dashed arrows: influences on costs.
Purple italics: factors amenable to direct intervention.
Black italics (population aging, cardiovascular event fatality): other specified trends.
Black nonitalics: all other variables, affected by italicized variables and by each other.
Page 2
VOLUME 7: NO. 1
JANUARY 2010
www.cdc.gov/pcd/issues/2010/jan/08_0231.htm • Centers for Disease Control and Prevention 3
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services, the
Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only and
does not imply endorsement by any of the groups named above.
costs associated with these events and with the identified
risk factors.
The model starts with conditions in the United States in
1990 and simulates them continuously through 2040. The
population without CVD and risk factor prevalence rates
are represented as dynamic stock or state variables, sub-
divided by sex and age group (18-29 y, 30-64 y, and ≥65 y).
Smoking and obesity are viewed as reversible conditions,
whereas diabetes, high blood pressure, and high cholester-
ol are viewed as chronic conditions that are not reversible
but that can be controlled, with the help of good-quality
primary health care, to reduce CVD risk.
The incidence of first-time CVD events in the model is
driven by the effect of several direct risk factors, based on
a widely used risk calculator from the Framingham Heart
Study (7). We modified that calculator in several ways
for this study, most fundamentally by estimating annual
risks at a population level on the basis of risk factor
prevalence rates, rather than at the level of an individual.
(A detailed description of the modified calculator is avail-
able at http://sustainer.org/cvd/documents/SELI_App1.
pdf). We also recognized the direct effect on CVD events,
especially myocardial infarctions, from secondhand smoke
and small particulate matter (PM 2.5) air pollution (8-11).
Furthermore, because deaths from CVD have declined
partly because of improvements in emergency and acute
care, we incorporated a downward trend in the CVD case-
fatality rate for 1990 through 2003 (12).
Obesity contributes to CVD, largely through diabetes,
high blood pressure, and high cholesterol levels (13). Other
indirect influences in the model are physical inactivity,
poor diet, psychosocial stress, and smoking as it affects
diabetes and obesity (14-21).
Both the direct and indirect influences in the model
may be modified by 19 interventions (Table 1). These 19
interventions could be implemented at a city, county, or
state level rather than requiring changes nationwide.
In functional terms, the interventions are of a few basic
types: those that provide broader access to health-promot-
ing services, those that promote desirable behaviors, and
those that tax or regulate to deter undesirable behaviors.
Cost calculation
We used a common metric — constant 2005 dollars — to
track medical and productivity costs (for morbidity and
mortality) that might be affected by the 19 interventions.
We measured the societal value of morbidity (sick days)
and premature mortality (years of life lost) using a human
capital approach, which estimates the market value of lost
productivity at work and at home (22). (A detailed descrip-
tion of cost calculations is available at http://sustainer.org/
cvd/documents/SELI_App2.pdf.) This summary of medical
and productivity costs can determine whether any inter-
vention, or package of interventions, is justified by its likely
aggregate consequences, or “total consequence costs. We
did not estimate the costs of interventions. However, the
total consequence costs can inform spending decisions. For
example, suppose that for a given intervention the model
calculates a total consequence cost savings of $50 per cap-
ita. Planners may then conclude that up to $50 per capita
could be justifiably spent to implement that intervention
and still create positive net benefits to society.
The model tracks 3 types of intervention consequences:
• Medical and productivity costs attributable to fatal and
nonfatal CVD events.
• Medical and productivity costs attributable to noncar-
diovascular complications of smoking (eg, lung cancer),
diabetes (eg, blindness), high blood pressure (eg, kidney
failure), and obesity (eg, colorectal cancer). We have thus
far been able to quantify these costs, but not yet the costs
related to noncardiovascular complications of stress (eg,
depression), physical inactivity (eg, back pain), or poor
diet (eg, colorectal cancer).
• Costs of services and products to manage risk factors.
These include medications and visits for managing
chronic disorders, mental health services, weight-loss
services, and smoking cessation services and products.
Model calibration
Although the model is meant to investigate interven-
tions in localities such as Austin/Travis County, we began
by calibrating it to represent the entire United States.
This approach enabled more precise estimation, given that
certain data were either unreliable or unavailable at the
local level. The results are generally reported as per capita
estimates to facilitate interpretation at a local level. Table
2 lists the major information sources on which the model
is based (23-32).
The model specifies initial (1990) incidence rates for
Page 3
VOLUME 7: NO. 1
JANUARY 2010
4 Centers for Disease Control and Prevention • www.cdc.gov/pcd/issues/2010/jan/08_0231.htm
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services, the
Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only and
does not imply endorsement by any of the groups named above.
smoking, obesity, diabetes, high blood pressure, and high
cholesterol, as well as cessation rates for smoking and
obesity. These parameters have been set so that the model
accurately simulates the observed changes in prevalence
rates in the National Health and Nutrition Examination
Survey from 1988-1994 to 1999-2004.
The model also contains 56 causal links requiring the
estimation of relative risks, effect sizes, or initial values.
Many of these parameters were estimated through the use
of published studies, meta-analyses, and in some instanc-
es, ad hoc surveys of veteran practitioners (33). Because
most of these parameter estimates have some level of
uncertainty, we also identified lower and upper bounds to
be used for sensitivity analysis.
Model testing
Having calibrated the model to accurately reproduce
observed trends in risk factor prevalence rates as well
as CVD events and deaths, we then explored plausible
futures. A base case scenario assumed no changes after
2004 in many of the local determinants of risk in the
adult population, including healthiness of diet, extent of
physical activity, stress, use of quality primary care, air
pollution, and the prevalence rates of smoking and obesity
among incoming 18-year-olds. This base case should not
be taken as a statement about what is most likely to hap-
pen in the absence of intervention, but rather serves as a
straightforward and easily understood benchmark against
which to compare intervention scenarios.
We tested interventions singly and in groups of similar
interventions. For all interventions, we assumed a 1-year
ramp-up during 2009, followed by full implementation
from 2010 through 2040. The significance of full imple-
mentation depends on the intervention, but in all cases is
based on effect sizes that the research literature or veteran
practitioners indicate should be possible:
• For the 7 marketing interventions and for taxes on
tobacco or junk food: doing the maximum that has been
demonstrated or seriously proposed somewhere in the
United States.
• For the 6 access interventions: raising access to 100%.
• For smoking restriction: reducing secondhand exposures
in workplaces and public places to zero.
• For air pollution: reducing small particulate matter by
50% from its 2001-2003 value.
• For sources of chronic stress: a 50% reduction.
• For the quality of primary care (ie, adherence to guide-
lines): improvement from a national average of 54% (27)
to 75%.
For each intervention scenario, we conducted separate
simulations using the midpoint, lower-end, and upper-end
values for all uncertain parameters. This method yielded a
range of plausible outcomes for each intervention scenario.
Discussion
Base case results
The base case projects that even after 2004, when we
assume no further changes to the model’s inputs, histori-
cal trends in the model’s risk factor prevalence rates will
continue through 2040, although with diminishing slopes.
In particular, the model projects further declines in smok-
ing (and, thus, secondhand smoke exposure) and high
cholesterol, and at the same time further growth in high
blood pressure and diabetes. The projected continuation of
past trends reflects the eventual death of older cohorts and
their replacement by younger cohorts with different habits
and characteristics. For instance, the continued decline
of smoking prevalence reflects the lower rate of smoking
among teens and young adults today than in previous
decades. Such demographic turnover also helps explain
the continued growth of high blood pressure and diabetes,
which occurs in the model as a legacy of the increase in
obesity a leading risk factor for both disorders from
1980 to 2004. The projected continuation of trends also
reflects the future aging of the population; the over-65
population will increase from 2010 through 2030. This
aging effect contributes to the projected decline in smoking
because smoking is much less common among the elderly.
It also contributes to the projected increase in high blood
pressure and diabetes because the prevalence of these dis-
orders is higher with increasing age.
Deaths from first-time CVD events, which declined by
35% from 1990 to 2004, are projected in the base case
to rebound by 33% from 2004 to 2040. Much of the past
decline is attributable to a 28% reduction in the event
fatality rate, from improvements in emergency and acute
care. But it also reflects an 11% decline in the rate of
CVD events that occurred, despite increases in high blood
pressure and diabetes, because of decreases in smoking,
Page 4
VOLUME 7: NO. 1
JANUARY 2010
www.cdc.gov/pcd/issues/2010/jan/08_0231.htm • Centers for Disease Control and Prevention 5
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services, the
Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only and
does not imply endorsement by any of the groups named above.
secondhand smoke, PM 2.5 air pollution, and uncontrolled
high cholesterol.
The potential future rebound in deaths anticipated by
our model reflects a 17% increase in fatality from CVD
events per capita and a 15% increase in the rate of CVD
events because of the aging of the population. Although
the base case projects no future increase in CVD events or
deaths within each age group, the aging of the population
will lead to an increase in the overall rate of CVD events
and deaths.
Per capita total consequence costs, which the model
calculates to have declined by 25% from 1990 to 2004, are
projected in the base case to decline by another 5% from
2004 to 2040. Total consequence costs encompass not only
CVD events (which account for 44% of the total costs in
2004) but also noncardiovascular complications of risk
factors (also 44%) and management of risk factors (12%).
Although per capita CVD event costs are projected to
increase by 12% from 2004 to 2040 (reflecting the increase
in the frequency of the events themselves) and per capita
risk management costs are projected to increase by 8%
(reflecting the growing demand for blood pressure and
diabetes treatment), these increases are more than offset
by a 25% decrease in noncardiovascular complications.
This decrease is due to the projected decline of smoking,
which in 2004 was responsible for more than 400,000
noncardiovascular deaths, primarily from lung cancer and
chronic obstructive pulmonary disease. These premature
non-CVD deaths from smoking account for a large fraction
(about 28% in 2004) of the total consequence costs calcu-
lated in the model.
Intervention scenario results
Individual tests of the 19 interventions suggest that
each can reduce deaths from first-time CVD events, and
most can reduce total consequence costs. Four of the
interventions, however, raise total consequence costs,
meaning that they increase risk factor management costs
more than they decrease the costs of medical events and
complications. These 4 interventions include the 2 that
encourage use of mental health services and the 2 that
encourage use of weight-loss services. However, because
of limitations in the model, planners should not dismiss
these interventions in the real world. In the case of mental
health services, we have not yet estimated the noncar-
diovascular costs of depression. In the case of weight-loss
services for obese people, our estimates of cost and benefit
are based on conventional dieting and exercise programs,
rather than on bariatric surgery, which, although more
costly, also appears to be more effective (34).
We present simulation results for only the 15 interven-
tions that in the model do not increase total consequence
costs (Figure 2). Such a multipronged approach may be
challenging to implement, given resource limitations, but
it is useful to look at what could be achieved.
The model suggests that if all risk factors in the model
were eliminated, the death rate could be reduced by approx-
imately 60% below the base case, which falls between the
50% to 75% rate that other authors have suggested (35).
This model dichotomizes blood pressure, cholesterol, and
diabetes as highor not highand does not further sub-
divide the “not highinto normal and borderline. Reducing
borderline conditions (prehypertension, borderline choles-
terol, prediabetes) to normal could further reduce CVD,
but we cannot explore this possibility with this model. (A
static analysis of the potential benefits of reducing both
high and borderline conditions is available at http://sus-
tainer.org/cvd/documents/SELI_App3.pdf.)
The model projects that a 15-component intervention
could reduce the first-time CVD event death rate rela-
tive to the base case by 20% (range based on sensitivity
analysis, 15%-26%) in 2015 and by 19% (range, 14%-25%)
in 2040. Thus, the interventions that could reduce CVD
deaths have a relatively rapid effect.
The effect of the interventions is more gradual with
regard to total consequence costs than it is with regard
to CVD deaths; nearly 40% of the eventual effect on
costs occurs after 2015 (Figure 2). If all risk factors in
the model were eliminated, consequence costs could
be reduced by approximately 80% below the base case.
Relative to the base run, the 15-component intervention
reduces consequence costs by 16% (range, 12%-23%) in
2015, eventually reaching 26% (range, 19%-33%) in 2040.
The reduction in consequence costs is $348 per capita
(range, $254-$514) in 2015 and $565 per capita (range,
$416-$722) in 2040.
The 15-component intervention may be better under-
stood by examining the incremental contributions of its
components grouped by topical cluster (Figure 3). We used
the same base case graph as in Figure 2 and then incre-
Page 5
VOLUME 7: NO. 1
JANUARY 2010
6 Centers for Disease Control and Prevention • www.cdc.gov/pcd/issues/2010/jan/08_0231.htm
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services, the
Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only and
does not imply endorsement by any of the groups named above.
mentally added the following topical clusters:
1) The 3 interventions that improve the use and quality of
primary care (Care).
2) The 6 interventions related to air quality and smoking
(Air).
3) The 5 interventions related to improved nutrition and
physical activity and the 1 intervention that would
reduce sources of stress (Lifestyle).
The relative effects of the clusters are different for CVD
deaths than they are for total consequence costs. Of the
3 topical clusters, the largest contributor to projected
reductions in CVD deaths, both in 2015 and 2040, is Care,
followed by Air and then a smaller (though growing) con-
tribution from Lifestyle. In contrast, the largest contributor
to projected cost reductions, both in 2015 and 2040, is Air,
followed by Lifestyle and then a smaller (and ultimately
negligible) contribution from Care. The contributions to per
capita cost reduction in 2015 are $235 from Air, $71 from
Lifestyle, and $42 from Care. The contributions in 2040 are
$393 from Air, $165 from Lifestyle, and $7 from Care.
Conclusions
The major factors that affect cardiovascular health at
a population level interact through causal pathways and
develop through gradual accumulations that defy simple
calculation. This dynamic complexity — and not just gaps
in data is a challenge for local leaders who want to
intervene most effectively given limited resources. Our
simulation model helps meet this challenge by integrating
what is known about the various risk factors in a single
testable framework for prospective policy analysis.
The simulations reported here point to several con-
clusions that local leaders and national allies may find
valuable.
1) The CVD death rate has declined in recent years, not
only because of improvements in emergency and acute care
but also because of reductions in the CVD event rate itself,
due to reductions in smoking, secondhand smoke, particu-
late air pollution, and uncontrolled high cholesterol. If this
progress does not continue at a similar pace in the future,
however, the CVD death rate will likely rebound strongly
as the population ages.
2) Medical and productivity costs associated with CVD
risk factors have declined because of declines in first-
time CVD events and consequent deaths, and because of
reductions in non-CVD deaths (especially lung cancer and
chronic obstructive pulmonary disease) associated with
Figure 2. Estimated impacts of a 15-component intervention, with ranges based on sensitivity testing, simulation model for cardiovascular disease (CVD) out-
comes. The 15 interventions are listed in Table 1 under the topical clusters of Care, Air, and Lifestyle.
Key:
Blue line = base case results.
Black line = expected reduction in death rate or costs from the 15-component intervention when the uncertain parameters are all set to their baseline values.
Orange shaded area around the black line = envelope of plausible outcomes in the 15-component intervention outcomes based on sensitivity testing. Upper
edge (least impact) results when all uncertain impact parameters are set to their lowest values, while lower edge (most impact) results when all uncertain
impact parameters are set to their greatest values.
Gray line = the model’s calculation of what the death rate or costs would be if all of the risk factors in the model — smoking, small particulate matter (PM 2.5)
air pollution, high blood pressure, high cholesterol, diabetes, obesity, poor nutrition, inactivity, and stress — were reduced to zero.
Page 6
VOLUME 7: NO. 1
JANUARY 2010
www.cdc.gov/pcd/issues/2010/jan/08_0231.htm • Centers for Disease Control and Prevention 7
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services, the
Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only and
does not imply endorsement by any of the groups named above.
smoking. Population aging will likely keep smoking preva-
lence on a path of decline into the future, so that even if
CVD deaths rebound, the total consequence costs need not
rebound.
3) Of 19 interventions that local planners may con-
sider for lowering CVD risk, at least 15 could reduce CVD
deaths without increasing total consequence costs.
4) Interventions aimed at reducing smoking and improv-
ing indoor and outdoor air quality can save lives relatively
quickly and can justify intervention spending equivalent to
as much as $300 per capita per year for 30 years (in 2005
constant dollars, without time discounting) to achieve the
full implementation targets. Most local health leaders are
already aware of the need for tobacco control and smoking
bans, but many may not be aware of the contribution of
particulate air pollution to CVD risk, even in areas like
Austin/Travis County without heavy pollution.
5) Interventions aimed at improving the use and qual-
ity of primary care to diagnose and control high blood
pressure, high cholesterol, and diabetes can save lives
quickly but should not be expected to save much on
total costs, primarily because of the high cost of medica-
tions. Consequently, the intervention spending to achieve
and maintain such improvement should not exceed the
equivalent of $25 per capita per year for 30 years. Other
researchers have similarly found that good preventive
care for chronic conditions may be cost-effective but is not
necessarily cost-saving (36,37).
6) Interventions to improve nutrition and physical activ-
ity and to reduce sources of stress take more time to affect
CVD deaths, as they gradually reduce obesity and other
chronic disorders. Nonetheless, their contribution grows
over time and may justify intervention spending equiva-
lent to as much as $100 per capita per year for 30 years.
The ability of particular localities to achieve full imple-
mentation within these cost limits may vary depending on
context and implementation factors. Potential extensions
and improvements to the model include the following:
• Modeling medical and personal costs for the post-CVD
event population and targeted interventions for sec-
ondary prevention to reduce the rate of recurrent CVD
events.
• Modeling the prevalence rates of borderline conditions
(prehypertension, borderline cholesterol, prediabetes)
and incorporating them in the CVD risk calculations.
• Modeling the prevalence of former smokers and incorpo-
rating their differential risks in the CVD event and cost
calculations.
Figure 3. Projected changes in the death rate from first-time cardiovascular disease (CVD) events and in total consequence costs per capita when 15 interven-
tions are combined, expressed in terms of clusters of interventions, simulation model for cardiovascular health outcomes.
Key:
Blue line = base case results.
Gray line = outcomes if all risk factors were reduced to zero.
Red line = implement the 3 interventions that improve the use and quality of primary care (Care).
Green line = add the 6 interventions related to air quality and smoking (Air).
Black line = add the 5 interventions related to improved nutrition and physical activity and the 1 intervention that would reduce sources of stress (Lifestyle).
This scenario includes all 15 interventions and is identical to the black line in Figure 2.
Page 7
VOLUME 7: NO. 1
JANUARY 2010
8 Centers for Disease Control and Prevention • www.cdc.gov/pcd/issues/2010/jan/08_0231.htm
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services, the
Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only and
does not imply endorsement by any of the groups named above.
• Incorporating the non-CVD consequences of stress,
physical inactivity, and poor diet.
• Estimating intervention implementation costs to better
inform intervention priorities.
• Incorporating additional independent risk factors for
CVD (eg, excess sodium intake, excess trans fat intake,
vitamin D deficiency, periodontal disease).
The model described here was created through a close col-
laboration with health planners in Austin/Travis County,
who are now using a locally calibrated version of the model
to support local strategy design and leadership develop-
ment. We plan to pursue similar engagements with col-
leagues elsewhere. With more widespread use, this tool
may help health planners across the country transform
local contexts to most effectively improve cardiovascular
health.
Acknowledgments
We acknowledge the contributions of Terry Pechacek,
Dave Buchner, Roseanne Farris, Parakash Pratibhu, Deb
Galuska, Adolfo Valadez, Karina Loyo, Rick Schwertfeger,
Cindy Batcher, Ella Pugo, Jessie Patton-Levine, Josh
Vest, Patty Mabry, John Robitscher, Alyssa Easton, Nancy
Williams, and Larry Fine.
Author Information
Corresponding Author: Jack Homer, PhD, Homer
Consulting, 4016 Hermitage Dr, Voorhees, NJ 08043.
Telephone: 856-810-7673. E-mail: jhomer@comcast.net.
Author Affiliations: Bobby Milstein, Darwin Labarthe,
Diane Orenstein, Centers for Disease Control and
Prevention, Atlanta, Georgia; Kristina Wile, Sustainability
Institute, Stow, Massachusetts; Justin Trogdon, RTI
International, Research Triangle Park, North Carolina;
Philip Huang, Austin/Travis County Health and Human
Services Department, Austin, Texas.
References
1. Bell J, Rubin V. Why place matters: building a
movement for healthy communities. Oakland (CA):
PolicyLink; 2007.
2. Labarthe D. Epidemiology and prevention of cardio-
vascular diseases: a global challenge. Gaithersburg
(MD): Aspen Publishers; 1998.
3. A public health action plan to prevent heart dis-
ease and stroke. Centers for Disease Control and
Prevention; 2003. http://www.cdc.gov/dhdsp/library/
action_plan/. Accessed August 14, 2009.
4. Sterman JD. Business dynamics: systems thinking
and modeling for a complex world. Boston (MA): Irwin
McGraw-Hill; 2000.
5. Homer JB, Hirsch GB. System dynamics modeling for
public health: background and opportunities. Am J
Public Health 2006;96(3):452-8.
6. Homer J, Milstein B, Wile K, Pratibhu P, Farris R,
Orenstein D. Modeling the local dynamics of cardio-
vascular health: risk factors, context, and capacity.
Prev Chronic Dis 2008;5(2). http://www.cdc.gov/pcd/
issues/2008/apr/07_0230.htm. Accessed August 14,
2009.
7. Anderson KM, Odell PM, Wilson PW, Kannel WB.
Cardiovascular disease risk profiles. Am Heart J
1991;121(1 Pt 2):293-8.
8. Pope CA 3rd, Burnett RT, Thurston GD, Thun MJ,
Calle EE, Krewski D, et al. Cardiovascular mortality
and long-term exposure to particulate air pollution:
epidemiological evidence of general pathophysiological
pathways of disease. Circulation 2004;109(1):71-7.
9. Dominici F, Peng RD, Zeger SL, White RH, Samet JM.
Particulate air pollution and mortality in the United
States: did the risks change from 1987 to 2000? Am J
Epidemiol 2007;166(8):880-8.
10. Mowery PD, Babb S, Bishop EE, Pechacek TF.
Comparison of two methods for estimating preva-
lence of exposure to secondhand smoke in the US
population. Paper presented at Towards a Smokefree
Society Conference; September 9-12, 2007; Edinburgh,
Scotland.
11. The health consequences of involuntary exposure to
tobacco smoke: a report of the Surgeon General. US
Department of Health and Human Services, Public
Health Service; 2006. http://www.surgeongeneral.gov/
library/secondhandsmoke/. Accessed August 14, 2009.
12. Morbidity and mortality: chart book on cardiovascu-
lar, lung, and blood diseases. National Heart, Lung,
and Blood Institute; 2007. http://www.nhlbi.nih.gov/
resources/docs/cht-book.htm. Accessed August 14,
2009.
13. Thompson D, Edelsberg J, Colditz GA, Bird AP, Oster
G. Lifetime health and economic consequences of obe-
Page 8
VOLUME 7: NO. 1
JANUARY 2010
www.cdc.gov/pcd/issues/2010/jan/08_0231.htm • Centers for Disease Control and Prevention 9
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services, the
Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only and
does not imply endorsement by any of the groups named above.
sity. Arch Intern Med 1999;159(18):2177-83.
14. Katzmarzyk PT, Gledhill N, Shephard RJ. The eco-
nomic burden of physical inactivity in Canada. CMAJ
2000;163(11):1435-40.
15. Fleshner M. Physical activity and stress resistance:
sympathetic nervous system adaptations prevent
stress-induced immunosuppression. Exerc Sport Sci
Rev 2005;33(3):120-6.
16. Elmer PJ, Obarzanek E, Vollmer WM, Simons-Morton
D, Stevens VJ, Young DR, et al. Effects of comprehen-
sive lifestyle modification on diet, weight, physical fit-
ness, and blood pressure control: 18-month results of
a randomized trial. Ann Intern Med 2006;144(7):485-
95.
17. Rozanski A, Blumenthal JA, Kaplan J. Impact of psy-
chological factors on the pathogenesis of cardiovascu-
lar disease and implications for therapy. Circulation
1999;99(16):2192-217.
18. Björntorp P. Do stress reactions cause abdominal obe-
sity and comorbidities? Obes Rev 2001;2(2):73-86.
19. Kouvonen A, Kivimaki M, Virtanen M, Pentti J,
Vahtera J. Work stress, smoking status, and smoking
intensity: an observational study of 46,190 employees.
J Epidemiol Community Health 2005;59(1):63-9.
20. Willi C, Bodenmann P, Ghali WA, Faris PD, Cornuz
J. Active smoking and the risk of type 2 diabe-
tes: a systematic review and meta-analysis. JAMA
2007;298(22):2654-64.
21. Flegal KM. The effects of changes in smoking preva-
lence on obesity prevalence in the United States. Am
J Public Health 2007;97(8):1510-4.
22. Haddix AC, Teutsch SM, Corso PS. Prevention effec-
tiveness: a guide to decision analysis and econom-
ic evaluation. 2nd edition. New York (NY): Oxford
University Press; 2003.
23. Thom T, Haase N, Rosamond W, Howard VJ, Rumsfeld
J, Manolio T, et al. Heart disease and stroke statistics
2006 update: a report from the American Heart
Association Statistics Committee and Stroke Statistics
Subcommittee. Circulation 2006;113(6):e85-151.
24. Mendez D, Warner KE, Courant PN. Has smoking
cessation ceased? Expected trends in the prevalence
of smoking in the United States. Am J Epidemiol
1998;148:249-58.
25. Sloan FA, Ostermann J, Conover C, Taylor Jr DH,
Picone G. The price of smoking. Cambridge (MA): MIT
Press; 2004.
26. Homer J, Milstein B, Dietz W, Buchner D, Majestic
E. Obesity population dynamics: exploring historical
growth and plausible futures in the US. Paper present-
ed at 24th International System Dynamics Conference;
July 23-27, 2006; Nijmegen, The Netherlands.
27. Asch SM, Kerr EA, Keesey J, Adams JL, Setodji
CM, Malik S, et al. Who is at greatest risk for
receiving poor-quality health care? New Engl J Med
2006;354(11):1147-56.
28. Russell MW, Huse DM, Drowns S, Hamel EC, Hartz
SC. Direct medical costs of coronary artery disease in
the United States. Am J Cardiol 1998;81(9):1110-5.
29. Sasser AC, Rousculp MD, Birnbaum HG, Oster EF,
Lufkin E, Mallet D. Economic burden of osteoporo-
sis, breast cancer, and cardiovascular disease among
postmenopausal women in an employed population.
Womens Health Issues 2005;15(3):97-108.
30. Flegal KM, Graubard BI, Williamson DF, Gail
MH. Cause-specific excess deaths associated with
underweight, overweight, and obesity. JAMA
2007;298(17):2028-37.
31. American Diabetes Association. Economic costs of dia-
betes in the US in 2002. Diabetes Care 2003;26:917-
32.
32. Clausen J, Jensen G. Blood pressure and mortality: an
epidemiological survey with 10 years follow-up. J Hum
Hypertens 1992;6:53-9.
33. Homer J. A dynamic model of cardiovascular risk:
reference guide for model version 8m. Atlanta (GA):
Centers for Disease Control and Prevention; 2008.
34. Sjostrom L, Lindroos AK, Peltonen M, Torgerson J,
Bouchard C, Carlsson B, et al. Lifestyle, diabetes, and
cardiovascular risk factors 10 years after bariatric
surgery. N Engl J Med 2004;351(26):2683-93.
35. Magnus P, Beaglehole R. The real contribution of
the major risk factors to the coronary epidemics:
time to end the “only-50%” myth. Arch Intern Med
2001;161(22):2657-60.
36. Russell LB. Preventing chronic disease: an important
investment, but don’t count on cost savings. Health Aff
(Millwood) 2009;18(1):42-5.
37. Kahn R, Robertson RM, Smith R, Eddy D. The impact
of prevention on reducing the burden of cardiovascular
disease. Circulation 2008;118(5):576-85.
Page 9
VOLUME 7: NO. 1
JANUARY 2010
10 Centers for Disease Control and Prevention • www.cdc.gov/pcd/issues/2010/jan/08_0231.htm
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services, the
Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only and
does not imply endorsement by any of the groups named above.
Tables
Table 1. Interventions Used in Simulation Model for Cardiovascular Health Outcomes, Organized by Topical Cluster
Topical Cluster Intervention
Care Access to affordable primary care
Promotion of primary care use
Good-quality primary care
Air Tobacco taxes and sales restrictions
Social marketing against smoking
Access to affordable smoking cessation services and products
Promotion of smoking cessation
Bans on smoking in public places
Regulations and incentives that reduce air pollution
Lifestyle Access to affordable healthy foods
Promotion of healthy diet
Junk food taxes and sales restrictions
Access to safe and affordable physical activity
Promotion of physical activity
Reduced sources of chronic stress through improved living conditions and social supports
Weight-loss and mental health
services
Access to affordable weight-loss services for the obese
Promotion of weight-loss services
Access to affordable mental health services for the chronically stressed
Promotion of mental health services
Page 10
VOLUME 7: NO. 1
JANUARY 2010
www.cdc.gov/pcd/issues/2010/jan/08_0231.htm • Centers for Disease Control and Prevention 11
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services, the
Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only and
does not imply endorsement by any of the groups named above.
Table 2. Information Sources Used in Simulation Model for Cardiovascular Health Outcomes
Topic Source
Population size, growth, and aging, and health care coverage US Census
Rates of cardiovascular events and deaths Reports from American Heart Association (23) and National Institutes of Health
(12)
Prevalence rates of smoking, obesity, and chronic disorders, and rates
of diagnosis and control of chronic disorders
National Health and Nutrition Examination Survey (NHANES), 1988-1994 and
1999-2004
Fraction of 18-year-olds who smoke, are obese, or have chronic disor-
ders
NHANES, Youth Risk Behavior Surveillance System
Prevalence of psychosocial stress Behavioral Risk Factor Surveillance System (BRFSS)
Access to and use of good nutrition, physical activity, and primary care BRFSS
Rates of smoking cessation Mendez et al (24), Sloan et al (25)
Rates of people moving from obese to nonobese Homer et al (26)
Trend in fraction of workplaces allowing smoking Surgeon General’s report (11)
Trend in small particulate matter (PM 2.5) air pollution Dominici et al (9)
Average quality of primary care Asch et al (27)
Medical costs, sick days, and years of life lost due to CVD events and
deaths
Social Security actuarial life tables, Haddix et al (22), Russell et al (28), Sasser
et al (29)
Noncardiovascular medical costs and sick days due to smoking, obe-
sity, diabetes, and high blood pressure
Linked files of Medical Examination Panel Survey, National Health Interview
Survey
Noncardiovascular mortality and years of life lost due to smoking, obe-
sity, diabetes, and high blood pressure
Centers for Disease Control and Prevention Smoking-Attributable Mortality,
Morbidity, and Economic Costs (SAMMEC), World Health Organization Statistical
Information System, Flegal et al (30), American Diabetes Association (31),
Clausen et al (32)
Page 11
  • Source
    • "In contrast, in selecting an agent-based model, the modeler is assuming that there is some emergent quality to the phenomenon-of-interest and that the underlying mechanisms explaining this are due to the micro-interactions between autonomous agents over time and between agents (that have the capacity to learn and adapt) and their environments (Cederman, 2005; Macy and Willer, 2002). Thus, questions about policy decisions and resources can be seen as most amenable to understanding through SDMs (e.g., Jones et al., 2006; Homer et al., 2010), whereas questions about the effects of social interactions and the built environment might require the micro-detail of agent-based models (e.g., Auchincloss and Diez Roux, 2008; Orr et al., 2014). However, it should be noted that some systems can be modeled using either approach and that hybrid simulations involving both approaches have also been developed in some areas of public health research, notably infectious disease epidemiology (Borshchev et al., 2007; Macal, 2010; Rahmandad and Sterman, 2008). "
    [Show abstract] [Hide abstract] ABSTRACT: As with all chronic diseases, it is now recognized that type 2 diabetes is a complex health issue, the etiology of which involves numerous risk factors operating at different ecological levels of analysis. However, this ecological complexity of the problem seldom manifests itself in the interventions for preventing the problem, which typically focus on changing behavior through universal health education, with the assumption of a homogeneous population. This paper examines the limitations of this way of framing the problem of type 2 diabetes, particularly its failure to capture the way in which this problem emerges because of dynamic interactions between individuals and their environments and how these interactions vary in fundamental ways depending upon the context within which they occur. Specifically, the paper examines how framing of type 2 diabetes in the Lower Rio Grande Valley (LRGV) affects which systems modeling method selects to understand the problem and to help guide policy-makers to ameliorate it. Each systems model has a paradigm characterizing it by a set of fundamental rules and underlying concepts. That is, each method bases on assumptions of how the model should be constructed and the knowledge obtainable from such assumptions. By assuming the model should be constructed in a certain way, the modeler (whether implicitly or explicitly) frames the problem by making assumptions about the phenomenon-of-interest. Choosing to develop any model asserts that the model proscribes to paradigmatic assumptions for how it would contribute something of value) in some capacity (for a purpose), which is ultimately affected by understanding, interpretation, and application of the problem. The paper describes how specific types of systems methods, those using agent-based models (ABMs) and system dynamics models (SDMs), can produce very different ways of understanding the problem of, and the leverage points for, type 2 diabetes in the LRGV. Additionally, it moves beyond simply outlining the general differences in the use and applications of ABM and SDM, to presenting models demonstrating how framing of the problem and model paradigmatic assumptions affect understanding of the problem of type 2 diabetes in the LGRV and its potential leverage points. While the examples are specific to a health problem in a specific community, the significance of such an approach is in its generalizability to how understanding social system behavior depends upon how framing the problem and the paradigmatic assumptions of the modeling method affect our understanding of social systems and public health problems.
    Full-text · Article · Jul 2015
  • Source
    • "SDM allows the user to remain committed to interventions that may make things worse before better (a common phenomenon ) and offers shorter-term expectations against which to track performance. Additional information on SDM and PRISM can be found elsewhere12131415. "
    [Show abstract] [Hide abstract] ABSTRACT: Dissemination and implementation (D&I) research seeks to understand and overcome barriers to adoption of behavioral interventions that address complex problems, specifically interventions that arise from multiple interacting influences crossing socio-ecological levels. It is often difficult for research to accurately represent and address the complexities of the real world, and traditional methodological approaches are generally inadequate for this task. Systems science methods, expressly designed to study complex systems, can be effectively employed for an improved understanding about dissemination and implementation of evidence-based interventions.
    Full-text · Article · May 2014 · International Journal of Behavioral Medicine
  • Source
    • "Additional topics on which we expect forthcoming publications include dissemination and implementation science , public health innovation, and mental health services research. In addition to the applied work presented in this volume, the application of systems science methods to specific research questions has included the Co-Guest Editors' own work on the use of system dynamics modeling for informing community level policy decisions to reduce cardiovascular disease (Homer et al., 2008; Homer et al., 2010; Loyo et al., 2013 ) and Markov modeling to inform tobacco control policy ( Levy, Mabry, Graham, Orleans, & Abrams 2010a, 2010b). We have also reviewed a variety of modeling applications in obesity (Levy et al., 2011). "
    [Show abstract] [Hide abstract] ABSTRACT: This supplement of Health Education & Behavior showcases the current state of the field of systems science applications in health promotion and public health. Behind this work lies a steady stream of public dollars at the federal level. This perspective details nearly a decade of investment by the National Institutes of Health’s Office of Behavioral and Social Sciences Research. These investments have included funding opportunity announcements, training programs, developing resources for researchers, cross-disciplinary fertilization, and publication. While much progress has been made, continuing investment is needed in the future to ensure the viability and sustainability of this young but increasingly important field.
    Full-text · Article · Oct 2013 · Health Education & Behavior
Show more