The Role of Behaviour C hange in Eating and Physical
Activity in the Battle against Childhood Obesity
Brian Dangerfield and Norhaslinda Zainal Abidin
Salford Business School,
University of Salford, Maxwell Building,
Salford M5 4WT, United Kingdom.
B.C.Dangerfield@salford.ac.uk N.ZainalAbidin@edu.salford.ac.uk
Obesity results from many influences including genetic and environmental. But eating and
physical activity are the two fundamental factors which influence obesity development. This work
investigates how food consumption choices and the extent of physical activity have an influence
on weight, BMI and the prevalence of obesity in a population of children aged 2-15 years.
Around forty years ago, the weight profile of the child (or adult) population was not particularly
abnormal. But with increases in energy-dense food consumption and reductions in physical
activity, weight, BMI and the prevalence of obesity has increased dramatically, especially since
the 1990's. Results from our model can explain these trends. If no interventions are taken to
counteract them then these markers will continue to increase in the future with consequential
health effects as the children become adults. In an effort to uncover the most effective
interventions, this study highlights the role that a system dynamics model can play to help
manage the problem in childhood, particularly through considering behavioural changes.
Optimisation experiments are conducted to determine whether interventions on the energy intake
or energy expenditure sides are likely to be most effective. In anticipated developments of this
research it is suggested that the Theory of Planned Behaviour could be used as a framework to
identify the motivational factors that influence children in their eating and physical activity
habits.
Keywords- System dynamics; childhood obesity; physical activity; food intake; average weight;
body mass index; obesity prevalence; energy balance.
1 INTRODUCTION
The prevalence of obesity among children is increasing, especially in developed countries like the
USA and the UK. In England, the prevalence of overweight and obese children has increased
since 1995. The split percentage of obese children aged 2 to 15 years in 1995 was11.1% for boys
and 12.2% for girls respectively, increasing to 16.1% for boys and 15.3% for girls in 2009 (NHS,
2011). If this trend continues, it is expected that around one quarter of the population aged below
20 could be obese in 2050 (McPherson et al. 2007).
Obesity is a complex issue because of two reasons: firstly, many potential factors contribute to
the development of obesity. For example, we can cite environmental factors (e.g. overlarge food
portions in fast food restaurants, the lack of neighbourhood sidewalks and food advertising),
genetic and family history factors, eating habits (e.g. higher consumption of fat and frequent
consumption of energy-dense take-away food) and the adoption of inactive lifestyles (Department
of Health, 2008; Dehghan et al. 2005; Kumanyika et al. 2002). Secondly, obesity evolves in a
non-linear fashion, with time delays and feedback processes making a contribution (Levy et al.
2010).
Even though diseases related to obesity in children are perhaps not as serious as those in adults,
obesity is part of a continuous process. Obesity in childhood underpins the development of
obesity in adults (Choudhary et al. 2007). Obese children have the strong possibility of becoming
obese adults when they are grown up (Speiser et al. 2005; Guo et al. 2002). Being an obese child
not only impacts on their health but also on their social life. Even more importantly, obese
children often exhibit low self-esteem and depression resulting from their physical condition.
They are stigmatised by society and often get bullied by their school peers. These phenomena
indirectly impact on their educational performance and their ability to learn (Datar et al. 2004;
Byrd et al. 2007).
An urgent solution is needed to prevent obesity continuing to increase in the future. There are
many negative impacts resulting from obesity such the costs incurred in the light of the health
consequences (NHS, 2011). Many of the previous studies of obesity have been of a qualitative
nature and were undertaken in a natural setting context. These type of studies allow researchers
to have direct observations of their subjects. Unfortunately, they are costly and time consuming.
Studies of a direct experimental nature, which aim to assess the impact on weight and BMI, have
not provided sufficient evidence of any significant immediate results and this suggests that more
time may be required in order to see any impact (Caballero et al. 2003).
Therefore, because of the complexities inherent in the obesity problem and with a combination of
time and cost constraints, it is necessary to research the utility that may be offered through
modelling methodologies. Simulation is one modelling method which is useful in helping to
evaluate large and complex problems (Ingalls et al. 2008; Taylor and Lane, 1998). By using
simulation models an evaluation of the effectiveness of a variety of policies can be made. The
randomized controlled trial settings found in most qualitative studies allow of only a limited
number of policy interventions to be considered as compared to settings where computer
simulation is employed (Levy et al. 2010). System dynamics modelling is a tool that provides a
framework to understand this complex health-related problem and is very useful for experiments
that involve non-linear effects and feedback processes (Sterman, 2000).
An initial purpose of this study is to develop a system dynamics model that can be employed to
quantify the relation between the energy balance, created by differences between energy intake
and energy expenditure, and how changes in this affect average weight in the child population,
together with its impact on BMI and the prevalence of obesity in the relevant age groups. Section
2 presents the model structure used to explain the overall obesity process taking into
consideration both food intake and physical activity behaviour. The discussion continues with
validation tests performed on the model in Section 3. To promote understanding of the ways in
which changes in the energy balance impact on weight, BMI and prevalence of obesity, Section 4
reports a sample of output runs from the base case model. Our modelling work is driven by a
most important objective: to assess policy interventions that provide workable solutions to the
problem. Interventions in this domain refer to population level modifications effected through
changes in eating and physical activity behaviour. Section 5 describes the results from two broad
illustrative interventions simulated with our model. Related to one of the model objectives in
identifying the most effective obesity prevention activities, Section 6 compares the results from
optimisation experiments to determine whether changes in eating or physical activity represent
the most effective approach to reverse current trends in average weight, BMI and prevalence of
obesity. Section 7 summarises the entire paper. Lastly, future work to be undertaken is reported
in Section 8.
2 MODEL STRUCTURE
This research is based on the common view that obesity in the UK (as elsewhere) occurs due to
changes in eating and physical activity associated with a modern lifestyle (WHO, 2011; Butland et
al. 2007; Kumanyika et al. 2002; Crawford and Ball, 2002). From an eating perspective, an
increase in energy intake is primarily driven by an increase in the average number of meals (eating
2
episodes), larger portion sizes and an increase in the consumption of fat associated with meals
taken outside of the home and particularly in fast-food restaurants (Young and Nestle, 2002;
Young and Nestle, 2007; Butte, 2000; Jéquier, 2001; Bowman and Vinyard, 2004). A marked
reduction in physical activity energy expenditure is related to the frequency, duration and intensity
of physical activity being replaced by sedentary behaviour (Kumanyika et al. 2002; National
Obesity Observatory, 2010; Office for National Statistics, 2011). A description of the conditions
that drive obesity are presented from a modelling perspective using the stock and flow diagrams
shown in figures 1-4.
As regards eating behaviour, the greater the portion size and number of meals taken results in a
larger consumption of total fat whereas the opposite condition applies for a lower consumption
(note the double-headed arrows on the flows in figure 1). The energy intake from outside meals,
associated with higher fat, carbohydrate and protein portion sizes, and the increasing number of
meals eaten outside the home results in a higher proportion of total energy intake being obtained
from this source. This process is reflected in figure 3 and by the + signs in figure 2. Turning to the
physical activity (PA) aspect, the higher the intensity, frequency and duration of light, moderate
and vigorous activity, the higher the PA energy expended. Conversely, if the frequency, intensity
and duration of sedentary behaviour were to increase, then PA energy expenditure decreases. This
process is illustrated by the + signs in figure 4.
Average of fit
portion size from
‘changes in fat protion| school
size from school meals,
Average of fit ‘Average of fit
a portion size from portion size from
‘changes in fat protion De Shen ‘changes in fat protion Dee email
size ffom home meals size ffom outside meal
N\A . 7
2 Av. fit consumption
Av number mak
‘Av number of meals consumed at school per
consumed outside per day day
Figure 1: Structure for modelling average daily fat consumption; a similar model structure is
developed for carbohydrate and protein portion sizes.
Fat conversion
Av. number of meals
consumed fiom outside
per day
Average of fit portion
size from outside
pp Energy intake ffom
ae + outside per year
Energy intake from’
ae os ie!
\
Average of protein \
portion size from outside Carbohydrate
conversion
Protein conversion
Average of carbohydrate
portion size fiom outside
Figure 2: Structure for measuring daily and annual energy intake from meals taken outside the
home; a similar model structure is developed for energy intake derived from home and school
meals,
Enerrgy intake fiom
home per year Energy intake fi
Sy _
Total yearly
euerny eke
fo
Energy intake from
school per year
+
Av. daily energy
intake
Days per year
Figure 3: Model structure to represent total energy intake
Vigorous
ffequency
Sedentary duration
‘Vigorous intensity
‘Vigorous duration —
Yearly physical actviy 4
>
_p» Increased physical Decreased phsyical
| =
|
Light inensiy |
Moderate duration
—— .
ffequency
\
Light duration
Figure 4: Model structure for measuring energy expended from physical activity
The basic model structure used for the overall obesity process is illustrated in figure 5. The aim
here is to hypothesise and synthesise ideas about the workings of the obesity process, taking into
account energy intake, energy expenditure, energy balance, weight and body mass index. A daily
positive energy balance is stored in muscle or fat. An increased body weight itself, however,
causes higher energy expenditure, specifically energy from the basal metabolism rate. An
equivalence between energy intake and energy expenditure (equilibrium in energy balance) will
ensure no weight change. The entire process can take years to show up as changes in the average
weight of the population (Homer et al. 2004). The model is calibrated in years and so changes in
daily energy balance and weight variability are smoothed out to result in a less sharp average
weight change in a year. However, because eating and performing physical activity is a daily
process, the model additionally computes energy intake and energy expenditure in daily units
(kcal/day) to aid understanding and to reflect common health nomenclature.
‘Thermic effect
of food
Energy intake
Av. physical
from home a ——
Energy intake froma . expenditure
school uae
~ wei
7
Av. daily energy
balance
Body Composition
2
Energy intake from
outside. mic
changes in av
weigh per day co
Ay. height
Figure 5: Interaction of energy intake, energy expenditure, energy balance, weight and
body mass index (adapted from Homer et al. 2004 and Abdel-Hamid, 2002).
The above description exhibits a conceptualisation which is rich in detail concerning age groups
and gender, the locations where food is consumed by children, the energy-giving components of
food and the various ways that energy can be burned; in particular different levels of physical
activity and the incidence of sedentary behaviour. Some models may exhibit a simpler structure
but we believe our formulation creates a richer policy space which allows more detailed
interventions to be explored. A high-level map of the current version of the model, reflecting the
above dynamic hypothesis, is included as an appendix.
Before these experiments can be conducted, we need to calibrate the model. In public health
research model calibration procedures can involve the use of literature data, simple calculations
and judgemental estimation (Taylor et al. 2005; Sterman, 2000). Data from the literature refers to
the use of parameter values obtained from relevant publications. Some of the parameters have
been obtained from important studies by Abdel-Hamid (2002 and 2003), Homer et al. (2004) and
Butte et al. (2007), as well as human physiology texts for known energy conversion coefficients.
The second approach involves simple calculations and relates to the use of simple statistical
estimations from the numerical data obtained from the Health Survey England from 1995 to
2009. The final approach of judgemental estimation has not been adopted here. This form of
estimation is done by guessing a value for a parameter and then running the model repeatedly
with a range of such trial values until a match is found between the real and observed behaviour
patterns.
3 MODEL VALIDATION
Validation tests have been carried out to gain confidence in our model structure and the
behaviour produced from the model. These tests include structure verification, parameter
verification, dimensional consistency, extreme condition, behaviour reproduction and behaviour
sensitivity. The choice of the tests employed is based on the purpose of the model (Sterman,
2000). For structure and parameter verification, the tests involve comparing how well the
structure and parameter values adopted match the real system. Both tests are completed by
reference to prior literature such as Abdel-Hamid (2002 & 2003), Homer et al. (2004 & 2006)
and Butte et al. (2007). Currently, the model contains 378 equations and mappings. Each of these
passes the dimensional consistency test. For the extreme condition test making constant the
parameters associated with energy balance and height results in the unchanging weight and body
mass index (BMI) behaviour demonstrated in Figure 12. This result equates with a condition
where energy balance is maintained over a lengthy period and a person is considered to be in a
steady-state or stable weight. In that situation energy intake and energy expenditure will be the
same (FAO/WHO/UNU, 2004). We now turn to the behaviour reproduction test. The output from
this test is presented in figures 8 and 10. Both figures demonstrate that behaviour patterns
produced from the model align with the reported historical data obtained from the Health Survey
England (1995 to 2009). The final test is a behaviour sensitivity test which is applied to explore
the sensitivity of model behaviour to variations in the parameters (constants) in the model. This
particular test can be performed in a number of ways and dynamic optimisation is one of the
possible methods. Details of the optimisation results are discussed in Section 6.
4 SOME BASE CASE RESULTS FROM THE MODEL
The work reported here has been influenced by previous system dynamics modelling studies
conducted by Abdel-Hamid (2002; 2003) and Homer et al. (2004; 2006). Their work has been
used as a guideline for our modelling work, although there are differences in model detail and, to
an extent, model purpose.
Although obesity occurs due to an increase in food intake and lack of physical activity (WHO,
2011; Butland et al. 2007; Kumanyika et al. 2002; Crawford and Ball, 2002), the approach used
in this research to model obesity has taken into consideration that obesity develops due to
feedback control between food intake, energy expenditure and fat stores as specified by Lambert
and Goedecke (2003), Homer et al. (2004) and Abdel-Hamid (2002; 2003). In an earlier paper
(Dangerfield and Abidin, 2010), we described the framework necessary to understand the
influence of energy intake (EI), energy expenditure (EE) and energy balance (EB) on the average
weight, BMI and prevalence of obesity in a population of children. This research attempts to
explore the interactions between EI, EE and EB. EB is the difference between EI and EE. Both
food intake and physical activity energy are investigated. A previous study by Egger and
Swinburn (1997) stresses that weight is gained because EI is greater than EE, EE less than EI or
both phenomena occur. In simple terms, weight gain is caused by eating more energy-giving
nutrients than energy burned. The opposite result of weight loss occurs if EI is less than energy
expenditure.
The following scenarios report on outputs from the base case model (Current). The aim is to
provide an understanding of the values and behaviour patterns exhibited by these variables (EE, El
and EB) together with their impact on the physical measurements and obesity prevalence in the
child population (aged 2-15 years).
Scenario 1: Here the objective is to investigate the impact of EI and EE on the EB and
consequently on average weight. If El is greater than energy expended, it creates a positive EB.
The illustrations are for the male age group aged 11-15 years, one of six age/gender groups
considered (see the appendix). Accumulation of a positive daily EB results in weight gain as
shown in Figure 6 below. Figure 7 illustrates the opposite scenario; here weight decreases due to a
negative EB.
Average daily energy intake, energy expenditure and weight
4,000. kcals/day
60 kg
nergy
balance
2,000. kcals/day
50 kg
0 kcals/day
40 kg
1970 1979 1988 1997 2006 2015 2024
Time (Year)
intake[Aged 11 10 15 years. Male} : Current kcats/day
fe eo . — ———n eee
[Aged 11 to'13 years,Male} : Current ke
Figure 6: Increase in average weight due to higher energy intake than energy expenditure
Average daily energy intake, energy expenditure and weight
2,000 kcals/day
60 kg
Negative energy
balance
— —
1,000 kcals/day
50 kg eS
0 kcals/day
40 kg
1970 1979 1988 1997 2006 2015 2024
Time (Year)
Av daily total energy intake[Aged 11 to 15 years,Male] : Current keals/day
“Av. daily total energy expenditure"[Aged 11 to 15 years,Male] : Current keals/day
“Av. Weight"[Aged 11 to 15 years,Male] : Current kg
Figure 7: Reduction in average weight due to higher energy expenditure than energy intake
Scenario 2: In this instance the objective is to demonstrate the impact of changes in average
weight and height on BMI. As average weight increases (decreases), whilst average height has
8
changed very little in the age groups over the sixty years considered, BMI increases (decreases)
in value. The BMI value is measured using the standard formula: BMI= weight/ (height
height). The units are in kg/metres-squared. We have tested this formulation in the model.
Figures 8 and 9 demonstrate the results obtained from the base case model. Both figures depict
that with increasing (decreasing) values for average weight and the little change in average
height, the BMI reflects a similar increasing (decreasing) value and behaviour pattern. For figure
8 the reported data for 11-15 year-old males is also included and the fact that the model trajectory
is within the broad range of reported data is encouraging.
Average weight, height and body mass index
40 kg/(metre*metre)
60 kg
1.8 metre
25. kg/(metre* metre) LLL 7 |
50 kg
1.4 metre
a
10 kg/(metre* metre)
40 kg
1 metre
1970 1979 1988 1997 2006 2015 2024
Time (Year)
BMI"[Aged 11 to 15 years,Male] : Current g/(metre*metre)
cg/(metre*metre)
kg
e] : Obesity Data [3] kg
fale] : Current metre
Height"[Aged 11 to 15 years,Male] : Obesity Data [3] metre
Figure 8: Increasing weight and BMI behaviour pattern
Average weight, height and body mass index
20 kg/(metre*metre)
60 kg
2 metre
15 kg/(metre*metre)
50 kg
1.5 metre
10 kg/(metre*metre)
40 kg
1 metre
1970 1979 1988 1997 2006 2015 2024
Time (Year)
"Av. BMI"[Aged 11 to 15 years,Male] : Current ———————_ kg /(metre* metre)
“Av. Weight" [Aged 11 to 15 years,Male] : Current kg
"Av. Height"[Aged 11 to 15 years,Male] : Current metre
Figure 9: Decreasing weight and BMI behaviour pattern
Scenario 3 Here the objective is to demonstrate the impact of BMI on the prevalence of obesity.
With increases (decreases) in the BMI value, the prevalence of obesity follows a similar behaviour
pattern. Figures 10 and 11 illustrate the behaviour produced for both cases: increasing (figure 10)
and decreasing (figure 11). Included in figure 10 is the reported data for males aged 11-15 years.
BMI and prevalence of obesity
40 kg/(metre*metre)
60 percentage
25 kg/(metre*metre)
35 percentage
10 kg/(metre*metre)
10 percentage
1970 = 1979 1988 1997 2006 2015 2024
Time (Year)
"Av. BMI" 11 to 15 years,Male] : Current kg/(metre*metre)
"Av. BMI" 11 to 15 years,Male] : Obesity Data [3] kg/(metre*metre)
Prevalence of obesity[Aged 11 to 15 years,Male] : Current percentage
Prevalence of obesity[Aged 11 to 15 years,Male] : Obesity Data [3] percentage
Figure 10: Increase in BMI results in an increased prevalence of obesity
10
40
60
25
30
BMI and prevalence of obesity
kg/(metre* metre)
percentage
kg/(metre* metre)
percentage
EE
kg/(metre* metre)
percentage
1970 1979 1988 1997 2006 2015 2024
Time (Year)
“Av. BMI"[Aged 11 to 15 years,Male] : Current kg/(metre* metre)
Prevalence of obesity[Aged 11 to 15 years,Male] : Current percentage
Figure 11: Decreasing BMI results in a decreasing prevalence of obesity
Scenario 4 A final scenario is reported to demonstrate an extreme case. If the height and EB is
indeed in balance throughout time (so that there is a zero energy gap), then the average weight and
BMI
20 kg/(metre* metre)
0.2 kcals/day
60 kg
15 kg/(metre*metre)
0.1 kcals/day
is also unchanged (figure 12).
Av. energy balance, height, weight and BMI
2 metre
1.5. metre
50 kg
10 kg/(metre*metre)
0. kcals/day
1 metre
40 kg
1970 1979 1988 1997 2006 2015 2024
Time (Year)
"Av. BMI"[Aged 11 to 15 years,Male] : Current kg/(metre*metre)
"Ay. daily energy balance"[Aged 11 to 15 years,Male] : Current kcals/day
"Av. Height"[Aged 11 to 15 years,Male] : Current metre
"Av. Weight"[Aged 11 to 15 years,Male] : Current kg
Figure 12: A constant energy balance yields a constant average weight and BMI
ll
The output runs from these base case scenario projections allow us to have confidence that the
model is producing results which are intuitive and which do not markedly differ from real-world
reported data, at least for the child age group we have been able to obtain data for.
5 EXPLORING BEHAVIOURAL CHANGES
Here we examine certain behavioural changes: specifically these reflect broad modifications in
eating habits and physical activity patterns and which are assumed to have taken place from the
baseline year of 1970. The objective here is to develop further confidence in the model as a tool
for investigating the consequences of such changes rather than (yet) simulating the effects of any
specific population level interventions which obviously could happen only after the present day.
As mentioned in the previous section, adverse changes in eating and physical activity behaviour
have contributed positively to the developing trends in obesity prevalence. Therefore we have
conducted two explorations in eating and physical activity patterns. For illustrative purposes the
graphs depict the results for 11-15 year old males.
Behaviour Change 1: Increasing fat intake to 40% while maintaining physical activity
In the first behaviour change, the aim is to examine the impact of eating behaviour on the average
weight, BMI and the prevalence of obesity. Increasing the proportion of fat in the diet to 40%, as
compared with the base case value (35.4%), clearly contributes to an increase in the total energy
intake. Compared to the base case run (figure 6), the total amount of energy intake is now changed
from 1857 keal/day to 1999 kcal/day in 1970 and, if this new fat proportion was maintained to
2030, it would reach 4169 kcal/day at that time (figure 13). The pattern of energy intake generates
an increasing trend over the simulated 60 year period because of the fractional growth rate in
portion sizes from outside, school and home meals. A similar increasing pattern is observed for
average weight and (very slightly) for energy expenditure. As long as a positive energy balance
(gap) is maintained (energy intake-energy expenditure), average weight will also follow an
increasing trend. Weight is shown to increase from 51 kg in 1970 to 52.8 kg in 2000 and is
predicted to increase again to 58.2 kg in 2030 if the revised fat proportion in the diet is maintained.
Figure 14 illustrates the linkage between weight and BMI. With the increasing average weight, the
BMI trend also increases over time. Average BMI is only 19.9 kg/metre® in 1970. The value
increases to 20.0 kg/metre’, 20.2 kg/metre’, 20.6 kg/metre’, 21.1 kg/metre’, 21.7 kg/metre? and to
22.6 kg/metre* in 1980, 1990, 2000, 2010, 2020 and 2030 respectively. A similar increasing
pattern is observed for the prevalence of obesity. The higher the average BMI value obtained, the
higher the prevalence of obesity, as would be expected. Figure 15 portrays this situation; the
prevalence of obesity, only 14% in 1970, is now projected to increase to 34.4% in 2030.
Energy intake, energy expenditure and weight
6,000 kcals/day
60 kg
3,000 kcals/day
50 kg
0. kcals/day
40 kg
1970 1979 1988 1997 2006 2015 2024
Time (Year)
Av daily total energy intake[Aged 11 to 15 years,Male] : Current keals/day
"Ay. daily total energy expenditure"[Aged 11 to 15 years,Male] : Curent. —————— A. kealls/day
"Ay. Weight"[Aged 11 to 15 years,Male] : Current kg
Figure 13: Increasing energy intake through increased fat consumption causes a rise
in average weight (energy expenditure more or less constant)
Weight, height and body mass index
40 kg/(metre*metre)
60 kg
2 metre
25 kg/(metre*metre)
50 kg
1.5 metre
10 kg/(metre*metre)
40 kg
1 metre
1970 1979 1988 1997 2006 2015 2024
Time (Year)
"Av. BMI" [Aged I] to 15 years,Male] : Current —————————_ kg /( mettre“ metre)
"Av. Weight" [Aged 11 to 15 years,Male] : Current kg
"Av. Height"[Aged 11 to 15 years.Male] : Current metre
Figure 14: Increasing weight associated with the experiment depicted in the previous
figure produces a linked increase in BMI
13
BML and prevalence of obesity
30 kg/(metre*metre)
70 percentage
15 kg/(metre*metre)
40 percentage
0 kg/(metre*metre)
10 percentage
1970 1979 1988 1997 2006 2015 2024
Time (Year)
"Av, BMI"[Aged 11 to 15 years, Male] : Current kg/(metre*metre)
Prevalence of obesity[Aged 11 to 15 years,Male] : Current percentage
Figure 15: Increasing BMI causes an increase in the prevalence of obesity
Behaviour Change 2: Maintaining food intake but increasing energy expended from physical
activity to 40% of the total
In the second behaviour change, the aim is to examine the impact of changes in physical
activity on weight, BMI and prevalence of obesity. Energy expended from physical activity is
increased to 40% of the daily total (base case= 13.2%) with commensurate reductions in the
percentages for the Basal Metabolism Rate and the Thermic Effect of Food to absorb the
change from 13.2% (see appendix). The renders the initial value for total energy expenditure to
be increased from 1858 kcal/day (base case) to 2688 kcal/day (intervention) in 1970. The
model output (figure 16) demonstrates that, from 1970 to 2001, average weight is reduced and
then it starts to stabilise (and even increase) between 2001 and 2030. The reduction in weight
occurs because energy expenditure is higher than energy intake and this has created a negative
energy balance/gap over the first 31 years. The slight increase observed in the final 29 years is
because the assumed change in the proportion of energy expenditure due to physical activity,
which changes the trajectory of the energy expenditure curve from the base case situation, is
still not sufficient to keep it above the base case (see figure 16) trajectory of energy intake for
the entirety of the run. Accordingly, there is a slight increase in average weight in the final 29
years of the run.
Energy intake, energy expenditure and weight
4,000 kcals/day
60 kg
2,000 kcals/day
50 kg
0. kcals/day
40 kg
1970 1979 1988 1997 2006 2015 2024
Time (Year)
Av daily total energy intake[Aged 11 to 15 years,Male] : Current keals/day
"Ay. daily total energy expenditure"[Aged 11 to 15 years,Male] : Curent ———-——————_______—. eals/day
"Av. Weight"[Aged 1 to 15 years,Male] : Current kg
Figure 16: Relationship between energy intake, energy expenditure and weight
With a reduction in average weight over the first 31 years of the run, BMI also reduces
similarly (figure 17), although again a slightly increase occurs after 2001. Average BMI
decreases from 19.9 kg/metre* in 1970 to 19.1 kg/ metre’ in 2000. The BMI value then
increases to 19.4 kg/metre” and 19.8 kg/metre’ in 2020 and 2030. A similar behaviour pattern is
observed in the prevalence of obesity (figure 18). The prevalence, 14% in 1970, decreases to
11.0% in 1990 and reaches 11.4% in 2020.
Weight, height and body mass index
25 kg/(metre*metre)
60 kg
2 metre
22 kg/(metre*metre)
50 kg eee
1.5 metre
19 kg/(metre*metre)
40 kg eee See
1 metre
1970 «1979 = 1988 1997 2006 2015 2024
Time (Year)
"Ay. BMI" [Aged 11 to 15 years,Male] : Current’ ———————_ kg/( metre“ metre)
"Av. Weight"[Aged 11 to 15 years,Male] : Current kg
"Ay. Height" [Aged 11 to 15 yearsMale] : Current metre
Figure 17: Relationship between average weight and BMI
BMI and prevalence of obesity
23 kg/(metre*metre)
35 percentage
21 kg/(metre*metre)
22.5 percentage
19 kg/(metre*metre) Pepsi ELL
10 percentage
1970 1979 1988 1997 2006 2015 2024
Time (Year)
“Av. BMI[Aged 11 to 15 years,Male] : Current kg/(metre*metre)
Prevalence of obesity[Aged 11 to 15 years,Male] : Current percentage
Figure 18: Relationship between BMI and obesity prevalence
Some conclusions may be drawn from the output graphs of the two interventions. These results
support the view that eating more food, particularly energy-dense foods with a high fat content, yet
maintaining physical activity patterns, results in an increase in average weight, BMI and obesity
prevalence. For the energy expenditure experiment the modification was effected through an
adjustment to the proportion of physical activity undertaken. With the energy expended from
physical activity increased to 40%, whilst maintaining the same food intake, a decrease in average
weight, BMI and obesity prevalence is the result, followed later by a rise for the reasons given
above.
6 IDENTIFYING THE MOST EFFECTIVE BEHAVIOURAL CHANGE USING
OPTIMISATION
Once confidence has been developed in terms of the model’s structure and behaviour, a series of
experiments are possible to help in the policy-making process. This work thus adds to the swathe
of research activity designed to highlight the most effective behavioural change strategies for
child obesity amelioration. Specifically, this section aims to make a comparison between changes
in eating and physical activity behaviour. Which is the most effective strategy for reversing
trends in child weight, BMI and prevalence of obesity? The results are presented from a total
population perspective (2-15 years) as a broad strategy to tackle obesity at the population level.
The population sector in the model allows the determination of relevant metrics over the full age
range by using weighted average calculations.
Dangerfield (2009) reiterates that there are two categories of optimisation in system dynamics:
policy optimisation and calibration optimisation. The first type called policy optimisation is
related to parameter changes which make improvements in model performance, for instance to
maximise sales revenue or to minimise costs. The other type, known as calibration optimisation,
involves the optimisation process searching for the parameter vector offering the best fit to a
given data series. This research applies the latter option: to identify the most effective
behavioural change strategies which achieve a desired weight target in 2020. However, since this
involves a future data series (as opposed to the more normal practice of fitting to past data in
calibration optimisation) the overall process is effectively a hybrid of policy and calibration
optimisations. The date of 2020 is important because the UK government has set a target to
reverse the obesity metrics back to 2000 levels by 2020 (Department of Health, 2008). This aim
can obviously be achieved through changes in average weight. Table | presents a plausible
average weight target needed by 2020 and figure 19 illustrates a graph for weight changes from
the baseline, starting in 2013, that would be compatible with UK government policy. The blue
line in figure 19 shows the trend for the total average weight from the simulated model (base
case) whereas the red line is a desired weight trajectory which passes through the target needed
by 2020 (=34.6 kg). Obviously the trajectory could be different with a slower initial decrease and
then a sharper fall to the target value in 2020. We have not yet explored other possible
trajectories.
Table 1: Changes in total average weight value for the past (1970-2010) and future (2020-
2030)
1970 1980 1990 | 2000 2010 2020 2030
Total average weight 36.69 38.56
(kg) (Baseline)
32.40 | 34.06 | 33.91 34.55 | 35.55
Desired total average 34.55 | 34.31
weight (kg)
Total av. weight (2-15 years)
40
35
20
1970 1976 1982 1988 1994 2000 2006 2012 2018 2024 2030
Time (Year)
"Total av. weight (2-15 yrs)" : Current
"Desired total av. weight (2-15 yrs)" : Current
Figure 19: Base case trajectory for total average weight (blue line) and a plausible desired
total average weight (red line) to achieve the target in 2020
In order to achieve the desired weight target of 34.6kg in 2020, behaviour changes must be made
in eating or physical activity or both. To be specific, the behavioural changes evaluated in this
research are achieved through two strategies which are: (1) improved food intake and (2)
increased physical activity with a concomitant reduction in sedentary behaviour. The changes can
only take place in the future so we have chosen 2013 for the policy changes to be rolled out.
Further, public policy changes cannot produce a sudden, step response so we have optimised on
rate of change parameters which are assumed to change in a gradual linear fashion over a period
of years. This is managed in the model using the RAMP function, one of the functions available
in the Vensim™ software. In order to attempt to achieve the necessary reduction in average
weight by 2020, the optimisation process was performed on rates of change in eating and
physical activity variables. It is reflected in the structure of equation | below. This generalised
equation can be applied to both eating and physical activity parameters, although the sign needs
to reflect whether a reduction or an increase is desirable. Changes in behaviour are assumed to
commence in 2013. The aim of the interventions is to encourage future child generations to adopt
healthy eating and physical activity behaviour and consequently improve their health in general.
Alterations made in eating and the physical activity parameters work by decreasing the amount of
EI and increasing both physical activity energy and total EE. Changes which occur in both EI and
EE result in changes in EB. The greater the reduction in EB, the more average weight is reduced.
Parameters=Initial value of parameters-RAMP (slope of change, 2013, 2030) ... (1)
18
The first experiment concerns changes made in eating behaviour. The baseline model (see figure
20) demonstrates that average daily energy intake increased between 1970 and 2030. The aim of
the eating optimisation is to investigate the impact of a reduction in the amount of energy intake
on average weight, BMI and the prevalence of obesity. In this experiment, changes are made in
two eating parameters while all physical activity parameters are held as they were in the base
case. There are seven parameters which are potentially useful for the experiment. However, we
chose only two parameters for the eating optimisation experiments. These are the fractional
growth rate in total meals consumed per year and the fractional growth rate in the fat portion
size for outside meals These two have the greatest influence on the determination of El.
Total av. daily energy intake (2-15 yrs)
4,000
3,250
keals/day
2,500
1,750
1,000
1970 1976 1982 1988 1994 2000 2006 2012 2018 2024 2030
Time (Year)
"Total av. daily energy intake (2-15 yrs)” : Current
Figure 20: Increased in energy intake between 1970 and 2030
The second experiment involves changes in activity behaviour. From the baseline model,
physical activity energy expenditure decreases between 1970 and 2030 as shown in figure 21.
Using optimisation, the aim is to increase the amount of physical activity energy from 2013 to
2030. There are seven possible parameters in the list of potential parameters to be optimised,
targeting on intensity and frequency. For the purpose of PA optimisation, only five parameters
are selected as follows: (1) metabolic rate involved in vigorous activity, (2) metabolic rate
involved in moderate activity, (3) fractional change rate in frequency per year of vigorous
activity, (4) fractional change rate in frequency per year of moderate activity, and (5) fractional
change rate in frequency per year of sedentary behaviour. Changes in all these parameters result
in beneficial changes occurring in the amount of physical activity energy expended.
Total av. daily physical activity (2-15 yrs)
400
350
300
keal/days
250
200
1970 1976 1982 1988 1994 2000 2006 2012 2018 2024 2030
Time (Year)
"Total av. daily physical activity (2-15 yrs)" : Current
Figure 21; Reduction in physical activity energy expended between 1970 and 2030
During the optimisation process, both eating and physical activity parameters are optimised for
the six age groups and gender (2-4 years, 5-10 years, 11-15 years, for both males and females).
From the eating optimisation, the average daily EI was reduced by means of a reduction in the
total meals consumed (eating episodes) per year and in the fat portion size of outside meals. By
contrast, the output obtained from the physical activity optimisation results in an increase in the
average daily physical activity energy expended and this is accomplished through an increased
frequency and intensity of moderate and vigorous activity and a decreased frequency of sedentary
behaviour. Incorporation of the optimum parameter values in the model each, separately, yields a
decrease in total average weight against the base case (not shown) ~ see figures 22 and 23.
The first finding from the eating (strategy 1) and physical activity (strategy 2) optimisation
experiments is that induced changes in the eating parameters offer the best means of a reduction
in average weight and provides the closest solution to the desired weight target in 2020. To be
more specific, the baseline results reveal that the total average weight in 2020 is 36.8 kg and after
eating optimisation, the total average weight in 2020 reduces to 35.6 kg. However, it does not get
below the target for 2020 until 2026. Similarly, the total average BMI is 19.7 kg/m’ at baseline
and this reduces to 19.1 kg/m? in 2020. Finally, the prevalence of obesity also exhibits a similar
reducing value from 24.0% in 2020 (baseline) to 21.1% in 2020 after the optimisation process.
In the physical activity optimisation, the total average weight in 2020 is 36.79 kg (baseline) and
this reduces to only 36.68 kg after optimisation. In 2030, the total average weight is 38.56 kg at
baseline, slightly reducing to 38.44 kg after optimisation. Similarly the BMI also exhibits a very
slightly reduced value: average BMI in 2020 is 19.70 kg/m? (baseline) decreasing to 19.69 kg/m*
after optimisation. Clearly the prevalence of obesity will mirror this fall: 24.04% in 2020
(baseline) against 24.02% after optimisation. The prevalence falls further by 2030 where in the
baseline the percentage is 28.75% reaching 28.51% after optimisation. A tabulated comparison of
both of the eating and physical activity optimisation experiments with the baseline is presented in
Table 2.
20
There is support for this finding in the public health literature. Skender et al. (1996), Ebersole et
al. (2008), Epstein et al. (2001) and Swinburn et al. (2009) all offer the view that changes in
modifying energy intake are more important compared to physical activity for obesity
amelioration. One possible explanation is there are no significant differences in physical activity
between obese and non-obese adults and children as suggested by Ebersole et al. (2008) and
Epstein et al. (2001). A second explanation is that although children can increase their physical
activity energy expenditure, they still continue to consume an increasing energy intake (El) at the
same time. This makes it harder for physical activity behavioural changes to achieve the required
amount of average weight reduction. By contrast, any beneficial reduction in energy intake
cannot be overcome by changes on the energy expenditure side because the changes underway in
reducing activity and increasing sedentary behaviour are not as fast-moving as those on the
energy intake side of the balance.
A second finding is that the achievement, in both average weight and prevalence, of the
government’s desired target is unlikely to be met by 2020. It may be possible by 2026 and can
only be achieved by adjusting eating behaviours. Physical activity changes alone will not allow
the target to be achieved at any point throughout the entire intervention period (2013-2030).
However, we express the caveat here that all these results derive from a limited set of parameter
combinations upon which we have experimented in the model. It may be possible to uncover a
combination which offers a significantly improved leverage, although our work thus far suggests
this is unlikely.
Total av. weight
40
35
i H
i 1
i \
i '
2 30 1 1
i 1
i f
i 1
| 2026=34.48kg i
25 | J t
1
H 2020=34.55kg i
H H
20 { i
1970 1976 1982 1988 1994 2000 2006 2012 2018 2024 2030
Time (Year)
“Desired total av. weight (2-15 yrs)" : Current
“Total av. weight (2-15 yrs)" : Strategy!: eating_optimisation
Figure 22: Comparisons of eating optimisation (red line) with the desired target (blue line)
for the total average weight between 2013 and 2030
21
Total av. weight
40
25
20
1970
1976 1982
Strategy
1988 1994 2000 2006 2012 2018 2024 2030
Time (Year)
Current
y2: PA_optimisation
Figure 23: Comparisons of physical activity optimisation (red line) with the desired target
(blue line) for the total average weight between 2013 and 2030
Table 2: Comparisons of total average weight, BMI and prevalence of obesity changes
resulting from two optimisation strategies
Total average weight Total average BMI Prevalence of obesity
(kg) (kg/m’) (%)
Baseline | Optimisation | Baseline | Optimisation | Baseline Optimisation
. 1970 2030 1970 2030 1970 2030
Baseline and|
intervention Baseline (Current)
strategies
32.40 38.56 17.98 20.61 12 28.75
Strategy 1: Eating optimisation
32.40 33.76 17.98 18.02 12 12.21
Strategy 2: Physical activity optimisation
32.40 38.44 17.98 20.55 12 28.51
22
7 SUMMARY
Eating more food and performing less physical activity are fundamental to obesity development
according to the scientific literature. The process is derived from the creation of a positive energy
balance when energy intake consistently exceeds energy expenditure. Output runs from the model
have supported the view that if children derive more of their energy intake from fat and burn less
energy from undertaking physical activity then population obesity metrics will turn adverse.
Although the base case changes simulated above have been of a broad general nature, the results
were encouraging enough to suggest that we could move forward and use the model to evaluate
specific targeted interventions to ascertain the future effects on the English child population.
The most important objective of this first strand of our interventions analysis has been to compare
changes in eating and physical activity behaviours in their role in childhood obesity amelioration.
The results reveal that a modification made in eating behaviour is the most effective tool in this
regard. The amount of average weight reduction from changes in eating behaviour is the closest
we came to hitting the desired average weight target to be achieved in 2020. However, we found
that the extent of change is still not enough to achieve the 2020 target. Therefore, it may be
necessary to expand the intervention period in order to achieve the weight and prevalence of
obesity target. This is true because the effects of a behaviour change process at the population
level will take a significant time to have an impact. Further, findings from this research will
deliver a message to policy makers, the food industry, practitioners and researchers of the
importance of collaboration in order to fully and effectively explore potential strategies for
reversing the obesity burden. This will help to achieve the UK government target of having a
healthy child population in the future.
8 FUTURE WORK
One of the facets we are striving to understand is: what are the most influential factors motivating
children to adopt either a healthy or a non-healthy lifestyle? Also, what can we say about the
effect on important metrics as a consequence of such changes ~ in either direction? Positive
behavioural change through an intervention surrounding these motivational factors would
represent a major milestone in overcoming current trends. Much of the current public health
research discusses the role of hard variables in physical activity and eating behaviour. Hard
factors in this context refer to quantifiable variables such as the intensity and frequency of
physical activity, for example. Turning to eating behaviours, examples include the number of
meals taken per day and the frequency of eating and snacking. For any model, behavioural
changes, involving soft factors, put more stress on the modelling technology and, in consequence,
the scope of insights which can be gained. It is easy to simulate the effects of changes in hard
variables (as we have done above) and many modelling methodologies would be capable of this.
But public health experts require to know how best to achieve those changes in the child
population. System dynamics is capable of incorporating soft factors into the model structure and,
as such, by exploring these softer dynamics it offers something above and beyond other
methodologies.
Behavioural change which concentrates on soft factors (e.g. attitude and perception) can also
contribute to uncovering and suggesting effective interventions. Owing to the importance of soft
factors as a contribution to influencing eating patterns and physical activity, further developments
in this research will explore the socio-psychological Theory of Planned Behaviour (TPB) due to
Ajzen (1991). TPB highlights the personal elements in the decision-making process and takes
into consideration the view that a person’s immediate surroundings shape their decision-making
and their attitudes.
23
Future work will include the incorporation of a developed model structure for understanding
physical activity and eating decision making processes, taking into account soft factors. Specific
behavioural change interventions can then be conducted to identify what are likely to offer the
most effective solutions to overcoming trends in obesity in the child population.
24
References
Abdel-Hamid, T. K. (2002) Modelling the dynamics of human energy regulation and its
implications for obesity treatment. System Dynamics Review, 18 (4), 431-471.
Abdel-Hamid, T, K. (2003) Exercise and diet in obesity treatment: an integrative system dynamics
perspective. Medicine & Science in Sports & Exercise, 35(3), 400-414.
Ajzen (1991). The theory of planned behaviour. Organizational Behaviour and Human Decision
Processes, 50: 179-211.
Bowman, S. A. and Vinyard, B. T. (2004) Fast food consumption of U.S. adults: impact on energy
and nutrient intakes and overweight status. Journal of the American College of Nutrition, 23(2),
163-168.
Butland, B., Jebb, S., Kopelman, P., McPherson, K.M Thomas, S., Mardell, J. and Parry, V.
(2007). Tackling obesities: Future choices-project report, 2“ edition. Report for Foresight,
Government Office for Science. Available from: _ http://www.bis.gov.uk/foresight/our-
work/projects/published-projects/tackling-obesities/reports-and-publications [Accessed 20
February 2009]
Butte, N. F., Christiansen, E. and Sorensen, T. I. A. (2007) Energy imbalance underlying the
development of childhood obesity. Obesity, 15(12), 3056-3066.
Butte, N. F. (2000) Fat intake of children in relation to energy requirements. The American
Journal of Clinical Nutrition. 72 (suppl), 1246S-1252S.
Byrd, J. (2007). The impact of physical activity and obesity on academic achievement among
elementary students. Available at http://cnx.org/content/m14420/latest/ [Accessed 20 November
2010)
Caballerro, B., Clay, T., Davis, S. M., Ethelbah, B., Rock, B. H., Lohman, T., Norman, J., Story,
M., Stone, E. J., Stephenson, L. and Stevens, J. (2003). Pathways: a school-based, randomized
controlled trial for the prevention of obesity in American Indian schoolchildren. American
Journal of Clinical Nutrition, 38, 1030-1038.
Choudhary, A. K., Donnelly, L. F., Racadio. J. M. and Strife, J. L. (2007). Diseases associated
with childhood obesity. American Journal of Roentgenology, 188: 1118-1130.
Crawford, D. and Ball, K. (2002) Behavioural determinants of the obesity epidemic. Asia Pacific
Journal Clinical Nutrition. 11 (Suppl), S718-S721.
Dangerfield, B. C. and Abidin, N. Z. (2010) Towards a model-based tool for evaluating
population-level interventions against childhood obesity. In: The Proceedings of 28" International
System Dynamics Conference, July 2010, Seoul, Korea, July 25-29, 2010.
Dangerfield, B. C. (2009) Optimisation of system dynamics models. In: Meyers, R. A. (Ed)
Encyclopaedia of Complexity & Systems Science, Springer, New York, USA. 9034-9043.
Datar, A., Sturm, R. and Magnabosco, J. L. (2004). Childhood overweight and academic
performance: National study of kindergartners and first grades. Obesity Research, 12:1, 58-68.
Dehghan, M., Akhtar-Danesh, N. and Merchant, A. T. (2005) Childhood obesity, prevalence and
prevention. Nutrition Journal, 4(24), 1-8.
25
Department of Health (2008) Healthy weight, healthy lives: A cross-government strategy for
England. Available from:
http://webarchive.nationalarchives.gov.uk/20100407220245/http://www.dh.gov.uk/en/Publications
andstatistics/Publications/PublicationsPolicyAndGuidance/DH_082378 [Accessed 10 May 2009]
Ebersole, K. E., Dugas, L. R., Durazo-Arvizu, R. A., Adeyemo, A. A., Tayo, B. O., Omotade, O,
O., Brieger, W. R., Schoeller, D. A., Cooper, R. S. and Luke, A. H. (2008) Energy expenditure and
adiposity in Nigerian and African-American women, Obesity, 16(9), 2148-2154.
Epstein, L. H., Gordy, C. C., Raynor, H. A., Beddome, M., Kilanowski, C., and Paluch, R. (2001)
Increasing fruit and vegetable intake and decreasing fat and sugar intake in families at risk for
childhood obesity. Obesity Research, 9(3), 171-178
Egger, G. and Swinburn, B. (1997). An ‘ecological’ approach to the obesity pandemic. British
Medical Journal, 315, 477-483.
FAO/WHO/UNU (2004) Human Energy Requirements. Report of a Joint FAO/WHO/UNU
expert consultation. Food and Nutrition Technical Report Series 1, Food and Agriculture
Organization, Rome.
Guo, S. S., Wu, W., Chumlea, W. C. and Roche, A. F. (2002) Predicting overweight and obesity in
adulthood from body mass index values in childhood and adolescence. American J ournal Clinical
Nutrition, 76, 653-658.
Homer, J., Milstein, B., Dietz, W., Buchner, D., and Majestic, E. (2006), Obesity population
dynamics: exploring historical growth and plausible futures in the US. Procs of 24" International
System Dynamics Conference, July 2006, Nijmegen, July 2006.
Homer, J., Jones, D., Milstein, B., Murphy, D., Essein, J. and Seville, D. (2004). Diabetes System
Model Reference Guidance. Revised Edition.
Ingalls, E.J., Kolesar, P. and Walker, W.E. (2008) Using simulation to develop and validate
analytic models: some case studies. Operations Research, 26 (2), page: 237-253
Jéquier, E. (2001) Is fat intake a risk factor for fat gain in children? The Journal of Clinical
Endocrinology & Metabolism, 86 (3), 980-983.
Kumanyika, S., Jeffery, R. W., MOrabia, A., Ritenbaugh, C. and Antipatis, V. J. (2002) Obesity
prevention: the case for action. International J ournal of Obesity, 26, 425-436.
Lambert, E. V. and Goedecke, J. H. (2003) Energy balance and energy expenditure in obesity- is
obesity a disease of inactivity? Sports Medicine, Clinical Review, 21-25.
Levy, D. T., Mabry, P. L., Wang, Y. C., Gortmaker, S., Huang, T. T-K, Marsh, T., Moodie, M.
and Swinburn, B. (2010). Simulation models of obesity: a review of the literature and
implications for research and policy, Obesity Reviews, 1-17.
McPherson, K., Marsh, T. and Brown, M. (2007) Tackling obesities: Future choices-modelling
future trends in obesity and their impact on health. Report for Foresight, Government Office for
Science. Available from: _http://www.bis.gov.uk/foresight/our-work/projects/published-
projects/tackling-obesities/reports-and-publications [Accessed 20 February 2009].
26
National Obesity Observatory (2010). Key data briefings, Available from:
http://www.noo.org.uk/NOO_pub/Key_data [Accessed on 31 May 2011]
NHS (2011) Statistics on Obesity, Physical Activity and Diet: England, 2011. The Department of
Health, The Information Centre. Available from: http://www.ic.nhs.uk/statistics-and-data-
collections/health-and-lifestyles/obesity/statistics-on-obesity-physical-activity-and-diet-england-
2011 [Accessed 15 Jun 2011]
NHS (2011) Health Survey for England-2009: Trend tables, The Information Centre, Available
from: http://www.ic.nhs.uk/statistics-and-data-collections/health-and-lifestyles-related-
surveys/health-survey-for-england/health-survey-for-england--2009-trend-tables [Accessed 10
April 2011]
Office for National Statistics (2011). Time Use Survey. Available from:
http://www.statistics.gov.uk/cci/nugget.asp?id=7. [Accessed 31 May 2011]
Speiser, . W., Rudolf, M. C. J., Anhalt, H., Camacho-Hubner, C., Chiarelli, F., Eliakim, A.,
Freemark, M., Gruters, A., Hershkovitz, E., lughetti, L., Krude, H., Latzer, Y., Lustig, R. H.,
Pescovitz, O. H., Pinhas-Hamiel, O., Rogol, A. D., Shalitin, S., Sultan, C., Stein, D., Vardi, P.,
Werther, G. A., Zadik, Z., Zuckerman-Levin, N. and Hochberg, Z. (2005). Consensus statement:
childhood obesity. The J ournal of Clinical Endocrinology & Metabolism, 90:3, 1871-1887.
Skender, M. L., Goodrick, G. K., Junco, D. J. D., Reeves, R. S., Darnell, L.,Gotto, A. M. and
Foreyt, J. P. (1996) Comparison of 2-year weight loss trends in behavioural treatments of obesity:
diet, exercise, and combination interventions, J ournal of the American Dietetic Association, 96
(4), 342-346.
Sterman, J. D. (2000), Business dynamics: systems thinking and modelling for a complex world.
Boston, MA: Irwin McGraw-Hill.
Swinburn, B., Sacks, G. and Ravussin, E. (2009) Increased food energy supply is more than
sufficient to explain the US epidemic of obesity. American Journal of Clinical Nutrition, 90,
1453-1456.
Taylor, K., Dangerfield, B. and Grand, J. L. (2005) Simulation analysis of the consequences of
shifting the balance of health care: a system dynamics approach. Journal of Health Services
Research & Policy, 10(4), 196-202.
Taylor, K. and Lane, D. (1998) Simulation applied to health services: some experiences so far
and opportunities for applying the system dynamics approach. Journal of Health Services, 3(4),
226-232.
World Health Organization (WHO) (2011) Obesity and overweight. Facts about overweight and
obesity Available from: (http://www.who.int/mediacentre/factsheets/fs311/en/). [Accessed on 20
May 2011].
Young, L. R. and Nestle, M. (2007) Portion sizes and obesity: responses of fast-food companies.
Journal of Public Health Policy, 28, 238-248.
Young, L. R. and Nestle, M. (2002) The contribution of expanding portion sizes to the US obesity
epidemic. American Journal of Public Health, 92 (2), 246-249.
27
Total Energy Intake Total Energy Expenditure
; Physical Activity Frequency of vigorous.
Number of Z tthe’ Ay quency of vigorous,
meals «eae netey snake Bom moderate, light and
- home sedentat
from home, s ry
school and Roamaaaveal Thermic Effect of Dura -_
; - ‘ood-derive : uration of vigorous.
outside Energy intake from Enerey tatake Energy Food (TEF) _ Vig s
school By Expenditure moderate, light and
, 5 : sedentary
Portion , size K_ | ee
rom ome, - —
outside and Energy intake from Basal Metabolism Intensity of vigorous,
school outside Rate (BMR) moderate, light and
sedentary
NG JN A
"ae
XN
-
Notations: This model is split into age group and gender ff
as follows: ENERGY BALANCE,
Body Composition
[2 to 4 years, Male]
[2 to 4 years, Female]
Prevalence of
Height :
[5 to 10 years, Male] Body Weight Obesity
[5 to 10 years, Female]
[11 to 15 years, Male] XL Body Mass
Index BMI) Obesity Prevalence (OP) Structure
111 to 15 vears. Femalel
\ Total population (2 to 15 years) z z
APPENDIX: High-level map showing the dynamic hypothesis underlying the model of childhood obesity and energy balance
28
