Anda di halaman 1dari 7

 OPPORTUNITIES AND DEMANDS IN PUBLIC HEALTH SYSTEMS 

Understanding Diabetes Population Dynamics


Through Simulation Modeling and Experimentation
Health planners in the Di- | Andrew P. Jones, MS, Jack B. Homer, PhD, Dara L. Murphy, MPH, Joyce D. K. Essien, MD, MBA, Bobby
vision of Diabetes Translation Milstein, MPH, and Donald A. Seville, MS
and others from the National
Center for Chronic Disease
Prevention and Health Pro- DIABETES MELLITUS IS A dynamics model developed to (normoglycemia); (2) prediabetes,
motion of the Centers for Dis- growing health problem world- explore the past and future bur- defined as having impaired glu-
ease Control and Prevention wide. In the United States, the den of diabetes—its morbidity, cose tolerance, impaired fasting
used system dynamics sim- number of people with diabetes mortality, and costs—in the glucose, or both4,5; (3) uncompli-
ulation modeling to gain a has grown since 1990 at a rate United States. Model develop- cated diabetes—that is, meeting
better understanding of dia- much greater than that of the ment was sponsored by the Di- the testing criteria for diabetes
betes population dynamics general population; it was esti- vision of Diabetes Translation but not yet symptomatic nor
and to explore implications mated at 20.8 million in 2005. and the Division of Adult and showing detectable signs of dis-
for public health strategy.
Total costs of diabetes in the Community Health at the Cen- ease in the eyes, feet, kidneys,
A model was developed to
United States in 2002 were esti- ters for Disease Control and Pre- or other organs; and (4) compli-
explain the growth of dia-
mated at $132 billion, with $92 vention (CDC). For background cated diabetes.
betes since 1980 and portray
possible futures through billion of that amount in direct on system dynamics methodol- The prediabetes and diabetes
2050. The model simulations medical expenditures and the ogy and applications, see Ster- (hyperglycemic) stages are fur-
suggest characteristic dy- other $40 billion in indirect costs man’s comprehensive textbook.3 ther divided among stocks of
namics of the diabetes pop- because of disability and prema- people whose conditions are di-
ulation, including unintended ture mortality.1 MODEL STRUCTURE AND agnosed or undiagnosed. Diagno-
increases in diabetes preva- There are no quick or easy CALIBRATION sis has dynamic significance be-
lence due to diabetes control, fixes for addressing the health cause it is a prerequisite for
the inability of diabetes con- and cost burdens of diabetes. Figure 1 displays the basic proper management and control
trol efforts alone to reduce di-
Like other dynamically complex causal structure of the system dy- of hyperglycemia and the often
abetes-related deaths in the
problems, diabetes is character- namics model. The full structure accompanying risk factors of hy-
long term, and significant de-
ized by long delays between also includes an inflow of adult pertension and hyperlipidemia;
lays between primary pre-
vention efforts and down- causes and effects, and the public population growth and outflows of and such management or control
stream improvements in health effort to address it is char- non–diabetes-related deaths. This can, in turn, greatly reduce the
diabetes outcomes. (Am J acterized by multiple concurrent structure reflects the knowledge rates of diabetes onset, progres-
Public Health. 2006;96:488– goals that may conflict with one and policy concerns of project sion, and death.6–10 In addition,
494. doi:10.2105/AJPH.2005. another. For example, although team participants and is grounded diagnosis affects the extent to
063529) planners have called for reduc- in the scientific literature on dia- which the prevalence of diabetes
tions both in the prevalence of betes, obesity, and related topics. in the population is recognized
diabetes and in deaths because Like all models, this one is a sim- and measured, as well as the
of its complications,2 the fact is plification: it omits many details amount of effort and money that
that fewer deaths, other things in order to enhance understand- are put into the clinical manage-
being equal, would lead to in- ing and includes assumptions that ment of prediabetes and diabetes.
creased, not decreased, preva- are uncertain to some degree. The Outside the population stock–
lence. Given such interconnec- model has evolved through a col- flow structure, Figure 1 shows
tions, a satisfactory solution will laborative and iterative process the potentially modifiable influ-
be found not in focusing on just that still continues. ences in the model that affect
1 aspect of the overall health At the core of the model is a the rates of population flow, in-
system—such as disease manage- chain of population stocks (ap- cluding influences that may be
ment, or detection, or risk factor pearing as boxes) and flows directly amenable to policy inter-
reduction—but rather in address- (appearing as double-thick ar- vention (indicated in italics).
ing all major components to- rows with valve symbols) por- These flow-rate drivers include
gether as a system. traying the movement of people prediabetes and diabetes detec-
We report results of simula- into and out of the following tion, prediabetes management,
tion experiments with a system stages: (1) normal blood glucose diabetes control, and (because of

488 | Opportunities and Demands in Public Health Systems | Peer Reviewed | Jones et al. American Journal of Public Health | March 2006, Vol 96, No. 3
 OPPORTUNITIES AND DEMANDS IN PUBLIC HEALTH SYSTEMS 

and topics are summarized in


Table 1.

BASELINE MODEL
BEHAVIOR

Figure 2 presents selected out-


put generated by simulating the
model through the historical pe-
riod starting in 1980 and then
into the future through 2050
under a set of hypothetical base-
line assumptions. The model re-
quires assumptions about the fu-
ture for each of the policy-related
model inputs indicated in italics
in Figure 1. In the baseline sce-
nario, we assume that no further
changes occur in obesity preva-
lence after reaching a value of
37% in 2006 (Figure 2a), and
that inputs affecting the detection
FIGURE 1—Overview of model structure, showing primary population stocks (boxes) and flows (arrows and control of prediabetes and
with valve symbols and cloud symbols for deaths), modifiable factors affecting flows (roman), and inputs diabetes remain fixed at their
amenable to policy intervention (italics). 2004 values through 2050. This
fixed-inputs assumption is not
meant to represent the project
its influence on the risks of pre- additional effects of changing ra- consequence of changes in racial team’s forecast of what is most
diabetes and diabetes onset) the cial and ethnic composition, as a and ethnic composition.) likely to occur to policy inputs in
population prevalence of obesity. Markov model by other re- The model’s parameters were the future, but it does make a
Prediabetes and diabetes detec- searchers does.11 However, by calibrated on the basis of histori- useful and transparent starting
tion may be improved by 2 including the effects of changes cal data available for the US point for policy analysis.
types of interventions: those in- in obesity prevalence, the system adult population, as well as esti- In addition to baseline simula-
creasing the glucose screening of dynamics model does capture mates from the scientific litera- tion output, Figure 2 also pre-
high-risk individuals by their what may be the most salient ture. The primary data sources sents historical data (Reported)
providers, and those increasing
access to preventive health care.
Diabetes control may be im- TABLE 1—Primary Data for Model Calibration
proved by 4 types of interven-
Information Sources Data Topics
tions: those enhancing clinical
12,13
management and those encour- US Census Bureau Population growth and death rates
aging patients to self-monitor Health insurance coverage
glucose levels, adopt healthy National Health Interview Survey 14 Diabetes prevalence
lifestyles, or use prescribed med- Diabetes detection
ications. (Another factor affect- National Health and Nutrition Examination Survey 15 Prediabetes prevalence
ing flow rates in the model, but Obesity prevalence
not indicated in Figure 1, is Behavioral Risk Factor Surveillance System16 Glucose self-monitoring
aging of the population, which Eye and foot examinations
affects death rates as well as Use of medications
prediabetes and diabetes onset Attending diabetes self-management classes
rates. The system dynamics Research literature Effects of disease control and aging on onset, progression, death, and costs
model, for the sake of simplicity, Direct and indirect costs of diabetes
does not explicitly depict the

March 2006, Vol 96, No. 3 | American Journal of Public Health Jones et al. | Peer Reviewed | Opportunities and Demands in Public Health Systems | 489
 OPPORTUNITIES AND DEMANDS IN PUBLIC HEALTH SYSTEMS 

increased significantly between


1980 and 2004. For example,
we estimate that the fraction of
primary care physicians who pe-
riodically test blood glucose lev-
els in their patients at high risk
for hyperglycemia rose steadily
from 69% in 1980 to 95% in
2004, and that such screening
has been the primary driver in
increasing the fraction of patients
with diabetes who have been di-
agnosed from 62% to 74% dur-
ing the same period.17 Model
simulation suggests that progress
on detection and management
has reduced the rate at which
people with diabetes move from
uncomplicated to complicated
diabetes, as well as the rate at
which people with complicated
diabetes die from the complica-
tions (Figure 2c).
From 1980 to 2004, the bene-
ficial influence of increased dia-
betes control managed to hold
mostly in check the harmful influ-
ence of increased disease preva-
lence: the model indicates that per
capita deaths from complications
Note. Reported obesity prevalence based on National Health and Nutrition Examination Survey,15 and reported diabetes prevalence based on
National Health Interview Survey.14 Baseline projection assumes that obesity prevalence rises to 37% in 2006 and remains fixed thereafter, and of diabetes decreased by about
that disease detection and control efforts all remain fixed after 2004. 5% (in fact achieving a 7% de-
cline by 2001 before giving back
FIGURE 2—Selected baseline model output, 1980–2050, and comparison to historical data for obesity
some of that gain from 2001 to
prevalence (a), diabetes prevalence (b), complication-related deaths per complicated cases (c), and
2004 because of some slowing
complication-related deaths (d).
in the rate of improvement in clin-
ical management apparent in the
on obesity prevalence in the costs of both urgent/extended the chain of causation seen in data16). This result occurred be-
overall population15 and diag- and preventive care of diag- Figure 1. Increased onset led to cause although the simulated prev-
nosed diabetes prevalence20 and nosed diabetes.17 increased prevalence, first of un- alence of complicated diabetes in-
shows how closely the simulated The 4 graphs in Figure 2 to- complicated diabetes and then of creased by 17% (Figure 2b) from
output for diagnosed diabetes gether tell the following story of complicated diabetes (Figure 2b). 1980 to 2004, the complications-
prevalence lies to the latter of diabetes prevalence and mortal- The second and opposing related death rate for people with
these time series. The model is ity for the historical period from force is a noteworthy improve- complicated diabetes decreased by
also able to reproduce available 1980 to 2004 as indicated by ment in the control of diabetes, 19% (Figure 2c) during the same
historical data on prediabetes model simulation. Two forces achieved through greater efforts time period.
prevalence, the diagnosed and have worked in opposition to to detect and manage the dis- The baseline simulation indi-
controlled fractions of people affect the number of diabetes- ease. It appears that glucose cates a future for diabetes preva-
with diabetes, population average related deaths. The first force is a screening and clinical manage- lence and diabetes-related deaths
BMI, obese fractions of people rise in the prevalence of obesity ment of diabetes by providers, as for the period 2004–2050 quite
with prediabetes and diabetes, (Figure 2a). This increase in obe- well as self-monitoring and adop- different from the past. With obe-
losses in health-related quality of sity led to a greater incidence of tion of healthier lifestyles by peo- sity prevalence fixed, by assump-
life because of diabetes, and the prediabetes and diabetes through ple with diagnosed diabetes, all tion, at its assumed high point of

490 | Opportunities and Demands in Public Health Systems | Peer Reviewed | Jones et al. American Journal of Public Health | March 2006, Vol 96, No. 3
 OPPORTUNITIES AND DEMANDS IN PUBLIC HEALTH SYSTEMS 

37% from 2006 onward, the dia- would rebound to surpass its assumed 20% in 2006 to 50% the flow rates of diabetes progres-
betes onset rate would be at its 1980 level by 2008. by 2012. Appropriate manage- sion and complications deaths.
high point as well, and diabetes ment of prediabetes includes These flow-rate reductions, in
prevalence would consequently INTERVENTION TESTS monitored regimens of in- turn, slow the growth in the num-
continue to grow through 2050 creased physical activity and ber of diabetes-related deaths
(Figure 2b). The rate of growth of What can be done now and in improved diet, plus medica- (Figure 3b). Because no further
diabetes prevalence would gradu- the future to reduce the number of tions for control of blood glu- improvement in clinical manage-
ally diminish, and prevalence deaths associated with diabetes cose, blood pressure, or lipids ment is assumed to occur after
would become more level (from complications? Simulation experi- as needed.8–10 2012, and because nothing has
about 2025 onward) only when ments with the system dynamics • Reduced obesity prevalence. The been done to slow the growth of
the outflow of deaths (because of model may help shed light on this obese fraction of the adult pop- diabetes prevalence (Figure 3a),
diabetes as well as all other causes) question. Here we consider just 3 ulation is reduced. Specifically, the rapid growth in complications
started to catch up with the inflow of many possible policy interven- this fraction is ramped down deaths resumes immediately after
of onset. The situation is compara- tion scenarios that may be tested from the assumed 37% in 2012. The resumed growth fol-
ble to the gradual filling of a bath- and compared with the baseline 2006 to 26% in 2017. This lows a trajectory that parallels
tub that has a slow drain—in this scenario. (A scenario consists of a reduction returns obesity prev- that of the baseline scenario, ac-
case, the drain being deaths of particular set of assumptions for alence to where it was in tually slightly exceeding it in
people with diabetes. In fact, with the future values of all time series about 1995. Real-life imple- terms of percentage growth from
the outflow of deaths being equal inputs in the model.) In each of mentation of this strategy 2012 to 2050. For this and any
to only about 4% per year of the these scenarios, a single policy-re- might involve consumer educa- other scenario (namely, scenarios
diabetes population (for example, lated input is changed starting in tion, insurance reimbursements involving improved self-monitoring,
in 2004 about 800000 deaths 2006 and ramping up through for calorie-control and physical medication use, or healthier
out of 20 million, about half of 2012 or 2017, remaining constant activity programs, and working lifestyles for people with diabetes)
these deaths because of complica- thereafter. The 3 scenarios are as with industry and government in which the proposed interven-
tions of diabetes), more than 20 follows: to bring healthier foods and tion has the sole effect in the
years would be required—after the improved opportunities for model of increasing the fraction
assumed peaking-out of obesity in • Enhanced clinical management of physical activity to a broader of diabetes patients who are con-
2006—for the growth in diabetes diabetes. The fraction of people spectrum of communities. trolled, the model suggests that as
prevalence (as a fraction of a grow- with diagnosed diabetes whose long as the controlled fraction is
ing adult population) finally to providers are adequately man- Resulting output graphs for 2 increasing, deaths from complica-
slow to a trickle. aging their disease (doing all variables—total diabetes preva- tions will grow more slowly;
With the prevalence of com- appropriate monitoring and ad- lence and per capita deaths from but after the increase in the con-
plicated diabetes growing by justment of medications) is in- complications—are shown in trolled fraction ceases, deaths will
38% from 2004 to 2050 (Fig- creased; specifically, this frac- Figure 3. The variables used resume a faster rate of growth in
ure 2b) and the death rate tion is ramped up from the in Figure 3 are the same as those line with the growth in diabetes
among the complicated cases de- baseline 48% in 2006 to 67% seen previously in Figures 2b and prevalence itself.
clining by only 2% (Figure 2c; by 2012. Real-life implementa- 2d but use narrower y-axis ranges Figure 3a indicates that the in-
this 2% decline reflecting some tion of this strategy might in- so that intervention impacts can tervention improving the clinical
continued reduction in the undi- volve broader adoption of clini- be seen clearly. For each of the 3 management of diabetes ulti-
agnosed fraction of complicated cal standards of care, better intervention tests, Figure 3 shows mately leads to a small but notice-
cases), deaths from complica- patient tracking systems, more how the intervention alters the able increase in the prevalence of
tions of diabetes would increase computerized reminder sys- behavior of the diabetes system diabetes. This is a direct reflection
on a per capita basis by 36% tems, and greater reimburse- from 2006 to 2050 relative to of the fact that deaths from com-
(Figure 2d). Absent further im- ments or other incentives for the baseline scenario. plications have been reduced rela-
provements in disease detection, the provision of preventive tive to the baseline scenario. Re-
management, and control, and clinical services. Enhanced Clinical turning to the bathtub analogy,
with obesity prevalence and dia- • Increased management of predia- Management of Diabetes the outflow drain has been made
betes onset remaining at their betes. The fraction of people As a result of this intervention, smaller whereas nothing has been
all-time highs, the past progress with diagnosed prediabetes the controlled fraction of the di- done to reduce the inflow. Just as
in mortality reduction would whose providers are adequately agnosed diabetes population in- the water in a bathtub with a
soon be undone; starting from its managing their disease is in- creases from 41% in 2006 to “backed up” drain rises further
lowest point in 2001, the per ca- creased; specifically, this frac- 47.5% by 2012. Increased con- than it would otherwise, one may
pita complication death rate tion is ramped up from the trol, in turn, immediately reduces say that the diabetes population

March 2006, Vol 96, No. 3 | American Journal of Public Health Jones et al. | Peer Reviewed | Opportunities and Demands in Public Health Systems | 491
 OPPORTUNITIES AND DEMANDS IN PUBLIC HEALTH SYSTEMS 

FIGURE 3—Model output for 3 intervention scenarios compared with the baseline scenario for diabetes prevalence (a) and complication-
related deaths (b).

becomes backed up when the prediabetes management inter- prevalence of prediabetes thus onset, as in the prediabetes sce-
death rate is reduced. vention is less than one might declines. Because there are fewer nario, but also reduces predia-
have hoped, it is still sufficient to people with prediabetes, and betes prevalence and avoids the
Increased Management of reduce deaths from complica- fewer of them are obese, dia- backing-up phenomenon seen in
Prediabetes tions, and is ultimately more ef- betes onset declines—by 15% to the prediabetes scenario. The
As a result of this intervention, fective at doing so than the dia- 19% relative to the baseline sce- model indicates that this dual ac-
many more people with diag- betes management intervention nario. This is enough of a decline tion is the key to the success of
nosed prediabetes are effectively described in the previous section. in onset to cause diabetes preva- the obesity reduction interven-
managed. Consequently, the per But it is not until after the year lence to peak in 2018 and then tion in stemming the growth of
capita rate of diabetes onset de- 2028 that per capita deaths decline continuously thereafter. diabetes prevalence and deaths.
creases (by about 5%), and re- under the prediabetes interven- Overall, diabetes prevalence rises
duced onset then leads to re- tion begin to dip below those only 5.5% from 2006 to 2050, CONCLUSIONS
duced prevalence. From 2006 to under the diabetes management compared with the 23.5% in-
2050, diabetes prevalence rises scenario. Also, it should be noted crease in the baseline scenario. The analyses presented in
by 17% under the intervention, that after 2028, although the The peak and decline of dia- this article indicate the sorts of
compared with a rise of 23.5% growth in per capita deaths is betes prevalence is ultimately insights and conclusions that
under the baseline scenario. Al- less under this intervention than translated into a similar peak and one may draw from simulation
though the reductions in diabetes under the baseline or diabetes decline in per capita deaths from experiments using the system
onset and prevalence are signifi- clinical management intervention complications. Per capita deaths dynamics model. In particular,
cant, they may be less than one scenarios, this growth in deaths under the obesity reduction sce- such experiments can improve
would expect from such a large does continue right through nario first dip below those under understanding of 4 characteris-
assumed increase in effective pre- 2050. Although the growth in the prediabetes management sce- tic dynamics of the simulated
diabetes management. This is be- diabetes prevalence has been nario in 2017 and first dip below diabetes population: (1) obe-
cause the intervention does not slowed under the prediabetes in- those under the diabetes man- sity’s role in driving the growth
do anything to reduce the onset tervention, it has not been agement scenario in 2021. The of prediabetes and diabetes
of prediabetes in the first place halted (Figure 3a). success of the reduced obesity prevalence; (2) the “backing up”
and thus, allows a “backing up” of intervention in halting and re- phenomenon—in which reduced
people in the prediabetes cate- Reduced Obesity Prevalence versing the growth of diabetes outflow from a population
gory. For many individuals, predi- As a result of this intervention, prevalence and complications stock causes a buildup in that
abetes management does not al- onset rates for prediabetes and deaths stands in contrast to the stock—that may undercut the
together prevent diabetes onset diabetes are reduced. Also, re- inability of the prediabetes and benefits of management and
but, rather, just postpones it. duced obesity allows more recov- diabetes management scenarios control efforts; (3) the inability
Although the reduction in ery from prediabetes back to to do so. Obesity reduction leads of management and control ef-
diabetes prevalence under the normal glycemic levels, and the to a lower flow rate of diabetes forts alone to reduce diabetes

492 | Opportunities and Demands in Public Health Systems | Peer Reviewed | Jones et al. American Journal of Public Health | March 2006, Vol 96, No. 3
 OPPORTUNITIES AND DEMANDS IN PUBLIC HEALTH SYSTEMS 

prevalence in the long term; and et al., unpublished data, 2006) experimentation help diabetes necessarily reflect those of the funding
(4) the significant delays be- suggest that the specified goal policy planners and other agencies.

tween primary prevention ef- for diagnosed prevalence reduc- stakeholders to better antici-
forts and downstream improve- tion may be virtually impossible pate the multiple effects of in- References
ments in diabetes outcomes. to achieve and moreover is terventions in both the short 1. National Diabetes Information
Clearinghouse Web site. National Insti-
Simulation experiments suggest inconsistent with other stated and the long term. tute of Diabetes and Digestive and Kid-
that these 4 characteristic dy- goals. ney Diseases. National Diabetes Statis-
namics in combination may System dynamics modeling tics. Available at: http://diabetes.niddk.
About the Authors nih.gov/dm/pubs/statistics/index.htm.
often cause intervention impacts could also conceivably be used Accessed November 30, 2005.
Dara L. Murphy is with the Division of
to look different in the short to integrate the effect of other Diabetes Translation, Joyce D. K. Essien 2. US Public Health Service. Healthy
term than they do in the long chronic disease programs with di- is with the Division of Public Health People 2010. Washington, DC: US Dept.
term. For example, in addition abetes prevention and control. Partnerships, and Bobby Milstein is with of Health and Human Services; 2000.
the Division of Adult and Community Available at: http://www.healthypeople.
to the experiments we have pre- One promising direction being Health, Centers for Disease Control and gov/document. Accessed January 26,
sented, we have also simulated pursued by the CDC is to develop Prevention, Atlanta, Ga. Joyce D. K. 2005.
strategies that represent a mix of a dynamic model of overweight Essien is also with the Center for Public
3. Sterman JD. Business Dynamics:
Health Practice at the Rollins School of
increased diabetes management and obesity capable of projecting Systems Thinking and Modeling for a
Public Health, Emory University, At-
Complex World. Boston, Mass: Irwin/
and reduced obesity prevalence. plausible alternative futures, al- lanta. Jack B. Homer is with Homer Con-
McGraw-Hill; 2000.
Comparing a mixed strategy to lowing an examination of a closer sulting, Voorhees, NJ. Andrew P. Jones
and Donald A. Seville are with the Sus- 4. Harris MI. Classification, diagnostic
one that focuses entirely on dia- look at the roles of nutrition and tainability Institute, Asheville, NC, and criteria, and screening for diabetes. In:
betes management, the experi- physical activity programs. An- Hartland, VT, respectively. National Diabetes Data Group, ed. Dia-
Requests for reprints should be sent to betes in America. 2nd ed. Bethesda, Md:
ments suggest that the focused other useful way to extend the
Dara L. Murphy or Kimbelian Barnes, National Institutes of Health; 1995:
diabetes management scenario work could be the development Division of Diabetes Translation, Centers 15–35.
may quickly reduce diabetes-re- of separate models of hyperten- for Disease Control and Prevention, Mail
5. Kahn R, Genuth S, Expert Commit-
lated complications and deaths sion and hyperlipidemia as well as Stop K-10, 4770 Buford Hwy, Atlanta,
tee on the Diagnosis and Classification
GA 30341 (e-mail: dlm1@cdc.gov;
but is less effective in the long explicit representation of them as of Diabetes Mellitus. Follow-up report
kbarnes1@cdc.gov).
on the diagnosis of diabetes mellitus.
term than the mixed strategy. risk factors (separate from obesity This article was accepted April 7,
Diabetes Care. 2003;26:3160–3167.
Such model-based insights though certainly affected by it) in 2005.
6. Diabetes Control and Complica-
may help the CDC and other or- the diabetes model. tions Trial Research Group. The effect
ganizations and individuals to Aside from such extensions, Contributors of intensive treatment of diabetes on
A. P. Jones managed the various aspects the development and progression of
identify more effective public more work remains in the refine-
of the modeling project, assisted with long term complications in insulin de-
health strategies and also to in- ment and testing of the existing modeling, and led the writing of the arti- pendent diabetes mellitus. N Engl J Med.
teract more effectively with one diabetes model and in identify- cle and design of the figures. J. B. Homer 1993;329:977–986.
another in diabetes planning ef- ing alternative future scenarios built the simulation model and collabo- 7. UK Prospective Diabetes Study
rated on the writing of the article. D. A. Group. Tight blood pressure control and
forts. The fact that the model is and intervention strategies suit- Seville synthesized the analysis that led risk of macrovascular and microvascular
an integrated tool interrelating able for simulation. The model’s to model design. B. Milstein improved complications in type 2 diabetes. Lancet.
all key dimensions of the burden assumptions, embodied in its the scope and structure of the model. 1998;352:703–713.
J. D. K. Essien framed and originated the
of diabetes should be helpful in equations and parameter esti- study, built institutional support for it, 8. Diabetes Prevention Program Re-
such endeavors. Although this mates, are continually being re- and created the evaluation framework search Group. Reduction in the inci-
for it. D. L. Murphy identified the use dence of type 2 diabetes with lifestyle
article has focused on the dy- fined as new information and intervention or metformin. N Engl J
and relevant insights of the model. All
namics of prevalence and deaths, ideas come to light. We are also authors conceptualized ideas, framed Med. 2002;346:393–403.
the model also generates mea- working to better specify the un- the structure of the model, and edited 9. American Diabetes Association
sures of morbidity and financial certainty surrounding parameter drafts of the article. and National Institute of Diabetes and
Digestive and Kidney Diseases. The pre-
costs and allows one to simulate values and performing sensitivity vention or delay of type 2 diabetes. Di-
how they too may be affected in analyses to determine the impact Acknowledgments abetes Care. 2002;25:742–749.
the future by alternative inter- of this uncertainty. Even in those This work was funded through the Asso-
10. Bowman BA, Gregg EW, Williams
ciation of Schools of Public Health Co-
ventions. cases in which the impact of operative Agreement and the Center for
DE, Engelgau MM, Jack L Jr. Translat-
ing the science of primary, secondary,
The system dynamics model the uncertainty may be great Disease Control and Prevention’s Divi-
and tertiary prevention to inform the
may also help in the setting of enough to affect policy conclu- sion of Diabetes Translation and Divi-
public health response to diabetes.
sion of Adult and Community Health, in
goals for diabetes management. sions, modeling may contribute collaboration with the Center for Public
J Public Health Manage Practice. 2003;
S8-S14.
Simulation experiments evaluat- by helping to prioritize empirical Health Practice at the Rollins School of
ing the national Healthy People research agendas. Public Health at Emory University, At- 11. Honeycutt AA, Boyle JP, Broglio
lanta, Ga (grant S3181). KR, et al. A dynamic Markov model for
2010 objectives for diabetes In summary, system dynam- Note. The views expressed in this ar- forecasting diabetes prevalence in the
(also see: Milstein, Jones, Homer ics simulation modeling and ticle are those of the authors and do not United States through 2050. Health

March 2006, Vol 96, No. 3 | American Journal of Public Health Jones et al. | Peer Reviewed | Opportunities and Demands in Public Health Systems | 493
 OPPORTUNITIES AND DEMANDS IN PUBLIC HEALTH SYSTEMS 

Care Management Science. 2003;6: Available at: http://www.census.gov/ Prevention, National Center for Health http://www.cdc.gov/brfss. Accessed Jan-
155–164. hhes/hlthins/historic/hihistt7.html. Statistics Web site. National Health uary 26, 2005.
12. Statistical Abstract of the United Accessed January 26, 2005. and Nutrition Examination Survey
17. Homer J, Jones A, Seville D. Dia-
States: 2002. Washington, DC: US 14. Centers for Disease Control and (NHANES). Available at: http://www.
betes System Model Reference Guide.
Census Bureau; 2003. Available at: Prevention, National Center for Health cdc.gov/nchs/nhanes.htm. Accessed
Hartland, Vt: Sustainability Institute;
http://www.census.gov/prod/www/ Statistics Web site. National Health In- January 26, 2005.
2004. Available at: http://sustainer.org/
statistical-abstract-02.html. Accessed terview Survey (NHIS). Available at: 16. Centers for Disease Control and pubs/diabetessystemreference.pdf. Ac-
January 26, 2005. http://www.cdc.gov/nchs/nhis.htm. Ac- Prevention, Behavioral Risk Factor Sur- cessed May 28, 2005.
13. US Census Bureau Web site. cessed January 26, 2005. veillance System Web site: Prevalence
Historical Health Insurance Tables. 15. Centers for Disease Control and Data and Trends Data. Available at:

494 | Opportunities and Demands in Public Health Systems | Peer Reviewed | Levy et al. American Journal of Public Health | March 2006, Vol 96, No. 3