The 34" International System Dynamics Conference, July 2016, Delft, Netherlands
MODELING THE LONG TERM DYNAMICS OF OBESITY AS A SOCIETAL
EPIDEMIC!
Gizem Aktas
gizem.aktas@ boun.edu.tr
Nefel Tellioglu
nefel.tellioglu@ gmail.com
Yaman Barlas
ybarlas@ boun.edu.tr
Hakan Yasarcan
hakan.yasarcan@ boun.edu.tr
Boazic¢i University, Industrial Engineering Department, Istanbul, Turkey
August 2016
ABSTRACT
Obesity has become a serious problem affecting people of all ages, racial and ethnic
backgrounds, and socioeconomic status, especially in the last decades. The rapid growth in
obesity represents a major public concern because of the negative health consequences of
obesity for individuals and society and the associated health expenses. The obesity problem is
largely due to the increase in the number of sedentary jobs, relatively high prices of healthy
food, and the increase in the accessibility of junk food. We constructed a population-level
system dynamics model and initialized/calibrated it using 1970's obesity data of US society.
For structural validation, we carried out extreme condition and sensitivity tests that yielded
positive results. For behavioral validation, we showed that the model output matches well the
real obesity data of US society between 1970 and 2010. We simulated the model for 40 more
years after 2010 and observed that the prevalence pattern of obesity continues to increase in
the near future as well. Finally, we tried three different scenarios with three different policy
interventions to alleviate the obesity problem in the future. We show that a compound
scenario that includes all three interventions may stop and revert the obesity prevalence.
Keywords: health modeling; obesity epidemic modeling; obesity factors; obesity in US; anti-
obesity policy; population models.
1 Supported by Bogazici University Research Grant no. 8880
1, INTRODUCTION
Obesity is defined as a disease resulted from the imbalance between energy intake and
expenditure. This imbalance is caused by a variety of factors such as genetic, environmental,
psychological, and socioeconomic (Madahian, 2012). In the last few decades, obesity has
become a worldwide epidemic problem and people from all ages, racial and ethnic
backgrounds, and socioeconomic status are affected by it. At a macro level, the problem is
largely due to the increase in urbanization and technological development (Cutler et al.,
2003). On the other hand, there are both genetic and social influences of adults on the child
obesity dynamics that in turn affects the adult obesity level of future. Such real life feedback
factors resulted in 22% and 13% increase in the adult and child obesity levels of US in the last
40 years, respectively (Ogden and Carroll, 2010; Fryar, Carroll and Ogden, 2014).
Even though the individual causes of obesity are recognized well, their effects on
the dynamics of the rapid increase of the obesity rate are not well understood. This may be the
reason why obesity level still has an increasing rate despite the variety of interventions.
Considering this need, our study and model takes into account the interactions of the real life
factors such as genetic, socio-economic, governmental, psychological and biological. We
build a population-level model, analyze the obesity dynamics as a societal epidemic and
discuss some policy interventions to fight it.
2. OVERVIEW OF THE MODEL
The target population is chosen from urbanized regions of US. Using these US data,
we construct a hypothetical region with a stable 1,000,000 population size. It is assumed that
there is no significant governmental concem about obesity in 1970-2000. At first, the model
has been initialized using the relevant initial parameters of 1970 to confirm the obesity
prevalence from 1970 to 2000. In 2000, the government intervention effects start operating.
We assumed that the government can have an impact on the healthy food prices, junk food
supply and junk food advertisement via taxes and other regulations.
In our conceptual model, there are four loops constructing the dynamic flows. A
simplified casual loop diagram is presented in Figure 2.1.
= Number of Obese
= Children
eS Children Food
Relative Healthy Consumption as,
Food Price ii i) 4
Children Obesity Loop Number of Obese
Profit of Junk Food
Industry
Rverage BMI of tn’
7 \ > Population
Extra Junk Food
Cy Consumption of Adults
Junk Food Supply Loop ~ +
4) Perceived BMX. imit
t Obesity Perception Loop °f Obesity!
Junk Food Supply
+ Junk Food
Accessibility
= Junk Food
Adv
Percentage
Governmental Regulation Loop
Figure 2.1: Simplified causal loop diagram
The main concem of above loops is the number of obese people in different age
groups of the society. Due to this concern, the model is constructed around 10 population
stocks created under three age categories, young children aging from 2 to 5, children aging
from 6 to 19, and adults aging from 20 to 60, and different obesity categories in each age
group, ranging from normal weighted to severely obese. A simplified stock-flow diagram is
shown in Figure 2.2.
Jroung Chiaren
| res
NOYC Births NOYC Aging 2
Obese Young Chisten NonObese Young
Nonotenng Rate KK) corn Obesing Rate
ores OYE Aging +
seve onan Crten J) (C9, nee Cen
Tenooeneg rae ECD Soon oe
Figure 2.2: The simplified stock-flow structure
3. MODEL BEHAVIOR
The model is simulated by using STELLA 10.0.4 software. The full time horizon of
the model is 80 years, which represents real time between 1970 and 2050. The base run of the
model depicts the dynamics of sub-population ratios of four different weight groups of adults
(see Figure 3.1). In Figure 3.2, the total obese ratios for the three age groups are plotted.
; __—
ae me
SS
17.
Se
| | ee
00
970,00 1900,00 Et 203000, 2080.09,
Figure 3.1: Behavior of the ratios of the four adult weight groups (Base Run)
In Figure 3.1, there is a significant decrease in the non-obese adult percentage
especially between the years 1990 and 2010. The overweight adult percentage stays almost
the same with a minor fluctuation from the beginning until the end of the base run simulation.
However, obese and severely obese adult percentages show a significant increase compared to
their initial values. The share of the extreme obese in total population sextuples itself at the
end of 80 year simulation. As the BMI values of the severely obese are quite high, this also
has associated health problems, societal problems, and costs. It is obviously important to
prevent such harmful trends. The governmental intervention loops that are mentioned in
Section 2 are active after the year 2000, which reflect on the dynamics of the percentages after
a delay. The impact of the governmental interventions can be observed in the dynamics of the
ratios after the first half of the simulation. As these interventions act against the feedback
loops that increase obesity over time, the increase dampens out.
: Adult Obesity Percent 2 Children Obesity Percent Young Children Obesity Percent
970,00 1990,00 2010.00 2030.00 2050.00
Figure 3.2: Total obese ratios in population for the three age groups (Base Run)
Starting with 14%, 5% and 5% obesity levels in year 1970, each age category of the
model reaches 47%, 26% and 15% in year 2050, after following an S-shaped behavior. The
reason for higher increase in years between 1990 and 2010 is the power of the positive
feedback loops described in Section 2. The explanation for the tapering of the rate of increase
around year 2000 is the damping effect of the governmental intervention loops.
4, MODEL VALIDATION
In order to validate the model, structural and behavioral validity tests are applied. As
structural validity test, extreme condition tests and sensitivity analysis are conducted.
Additionally, to analyze the behavior validity, behavior of the model for the time interval of
1970 — 2010 is compared to the real data.
Prevalence of Adult Obesity (Age 20-60) Prevalence of Child Obesity (Age 6-19)
Py x0
0 Pa
“0 0
#3 ais
0 10
10 5
° °
1970 1980 1990 2000 2010 2020 2030 204 2050 1970 198 1990 2000 2010-2020 2080 208 2080
Prevalence of Young Child Obesity (Age 2-5)
~@Real Data
—= Model Results
1970 1980 1990 2000 2010 2020 2030 208 ©2050
Figure 4.1: Prevalence of obesity within different age groups
[data is obtained from Ogden & Carroll (2010) and Fryar, Carroll & Ogden (2014)]
In Figure 4.1, we present real US data and model generated output in a comparative
way for behavioral validation. The simulation results for the time interval of 1970 — 2050 are
given, where the future predictions generated by the model can also be seen.
5. SCENARIO ANALYSIS
In the base model, government intervention loop starts from year 2000 and becomes
more intense as the obesity ratios increase. However, the base run of our model reveals that
the current moderate intervention by the government is not strong enough to prevent almost
half of the population becoming obese in year 2050. Therefore, we have tested stronger
governmental interventions for scenario analysis. First, we assume a more responsive
intervention by the government in reducing taxes on healthy food. Secondly, we assume a
more responsive intervention in increasing taxes on the junk food supply. Thirdly, we assume
a more responsive intervention by the government in limiting the advertisement of junk food.
However, the most effective one was to have a combination of all the three scenarios together.
Asaresult, adult obesity ratio goes to around 22% (see Figure 5.1).
“Adu Obesiv Percent 2 chiaven Ovesity Percent 2:Youne Chilten Ones Percent
eee sonny
,
—
Ls |
ol pm
1
oy
970,00 1990,00 201000 2020,00 72080,00
Figure 5.1: Results
(Three strong policy interventions applied together)
6. CONCLUSION
We constructed a population-level system dynamics model of obesity dynamics. The
model is initialized and calibrated using 1970’s obesity data of US society. Using these US
data, we construct a hypothetical region with a stable 1,000,000 population size. The obesity
problem is hypothesized to be influenced primarily by the number of sedentary jobs,
relatively high prices of healthy food, and the increase in the accessibility of junk food. There
are three major reinforcing loops involving these obesity factors that play critical roles in
obesity prevalence. The model also has a government intervention loop, an anti-obesity
control loop that is activated after year 2000. There are three age groups and four weight
(BMI) categories in the model. For structural validation, we carried out extreme condition
tests and sensitivity tests. For behavioral validation, we showed that the model output matches
well the real obesity data of US society between 1970 and 2010. We simulated the model for
40 more years after 2010 and observed that the prevalence pattem of obesity exists in the
future as well. This means that the current government intervention loop is not strong enough
to overcome the obesity reinforcing loops. In scenario analysis, alternative, stronger policy
interventions are tested. As a further research, we will continue to analyze different scenarios
and intervention policies, including their realism. There is also a potential value in turning the
model into an interactive policy game and make it available for experimentation by policy
makers and other stakeholders.
REFERENCES
Burke, M. A., F. Heiland, 2007, “Social Dynamics of Obesity”, Economic Inquiry, Jul
2007 45.3 ProQuest Central pg. 571.
Cutler D, E. L. Glaeser, and J. M. Shapiro, 2003, “Why Have Americans Become More
Obese?”, J ournal of Economic Perspectives, Vol. 17, No: 3,Summer 2003, pp. 93-118.
Fryar, C., M. D. Carroll, and C. L. Ogden, 2014, “Prevalence of Overweight and Obesity
Among Children and Adolescents: United States 1963-1965 Through 2011-2012”,
Division of Health and Nutrition Examination Surveys.
Homer, J., 2006, “Obesity population Dynamics: Exploring Historical Growth and Plausible
Futures in the U.S.”, 24th International System Dynamics Conference.
Madahian et al., 2012, “System Dynamics modeling of childhood obesity”, BMC
Bioinformatics, 13, Suppl 12, A13.
Ogden, C. and M. Carroll, 2010, “Prevalence of Overweight, Obesity, and Extreme Obesity
Among Adults: United States, Trends 1960-1962 Through 2007-2008”, Division of
Health and Nutrition Examination Surveys.
