Using System Dynamics to model the HIV/AIDS epidemic
in Botswana and Uganda
R. Howell’, O. Wesselink”, E. Pruyt
Delft University of Technology
Faculty of Technology, Policy and Management
P.O. Box 5015, NL-2600 GA Delft, The Netherlands
16 August 2013
1R J .Howell@ student.tudelft.nl, 7O.J.H.Wesselink@ student.tudelft.nl
Abstract
Uganda and Botswana present two interesting and contrasting cases in the AIDS epidemic.
System dynamics models of the AIDS epidemic in Botswana and Uganda were created to
examine the future development of the virus in both countries and evaluate existing and future
policy measures. The effect of existing and new policies such as abstinence only, educational
campaigns, increased availability of anti-retroviral therapy (ARTs) and condoms, and
circumcision were tested on key social and economic indicators. The most effective policy in
both countries is a combination of scaling up prevention of mother to child transmission
(PMTCT), condom use and availability, circumcision, and ART coverage. The epidemic in
Botswana seems to have leveled out; therefore, alternative policies such as circumcision will
be necessary to fight HIV in the future. In Uganda there is great need for improvement in
dealing with the AIDS epidemic. The model shows that HIV prevalence will increase without
drastic policy measures.
Key Words: Public health, public policy, economic development, Sub-Saharan Africa
1 Introduction
1.1 Problem Description
In June 2001 Botswana president Festus Mogae when speaking about AIDS/HIV stated: “We
are threatened with extinction. People are dying in chillingly high numbers. It is a crisis of
the first magnitude."
Sub-Saharan A frica has some of the highest HIV/AIDS rates in the world, a problem that is
decreasing life expectancies, leaving children without parents, and placing significant
economic burden on already developing economies. As the world watches as the tragedy of
AIDS unfolds and ravages countries already often plagued with disease, corruption, poor
infrastructure and health services, Uganda and Botswana present two interesting and
contrasting cases in the AIDS epidemic. Botswana has the most stable and transparent
government in A frica, but also one of the highest HIV rates in the world at almost 25% (UN
AIDS). It is unclear why Botswana’s HIV prevalence is so high but since the prevalence rate
is a percentage of the population, its high magnitude could also be attributed to Botswana’s
small population. Unlike most African countries Botswana has been open about the AIDS
problem and created aggressive policies to both reduce the number of infections and increase
the life expectancy of those living with AIDS. While these policies have dramatically reduced
the HIV prevalence, it has created a financial burden to the Botswana government. In
Botswana's National Development Plan 10 (NDP 10) report, outlining government goals
through 2016, the AIDS epidemic was a large focus. In NDP 10 Botswana has set targets of
annual GDP growth rates, targets that in a 2012 progress report have not been reached. The
AIDS epidemic has decreased both the productivity of the workforce in Botswana and the
number of working adults (also a result of the reduced life expectancy), factors that may
influence Botswana’s future economic growth . Although Botswana has achieved universal
coverage of anti-retroviral therapy (ART) (ie: more than 80% of the population having access
to ARTs), they have the challenging 2016 target of zero new HIV infections (NDP 10
Report). Since Botswana already has wide coverage of ARTs, high acceptance and use of
condoms, and a relatively low birth rate, it is predicted that the HIV prevalence rate will
simply level out but not exhibit the dramatic decrease experienced at the beginning of the
century (Avert, 2012).
Uganda has a much lower AIDS rate at 6.7% (UN AIDS) than Botswana but a different set of
problems. As the country where the first HIV case was discovered, Uganda enjoyed early
success in fighting the AIDS epidemic. It is unclear however whether their early decrease in
HIV prevalence was actually due to policy measures taken or due to outside factors such as
deaths due to AIDS and civil war. Currently, Uganda has one of the highest population
growths in the world, a phenomena which could increase the AIDS epidemic without policies
directed at reducing the number of new infections. Over the last ten years the HIV prevalence
rate has stayed constant/increased in Uganda. Additionally, although nearly 50% of Uganda’s
population is under 18, Uganda’s government is encouraging the rapid population growth in
the country, stating that the growth will be the country’s strength rather than addressing the
potential scale up of HIV infections resulting from the population increase (Avert, 2012).
Much of the infected population lacks access to anti-retroviral therapy, the relatively easy
treatment for mother to child transmission of AIDS (MTCT) is lacking, and wide acceptance
and availability of condoms is absent. Uganda’s lack of coherent policies regarding family
planning and AIDS is a large cause for concen within the development community.
Presently, Uganda spends a minimal amount of government spending on the AIDS epidemic;
most spending is from international organizations (UN AIDS). Modeling the dramatic
population growth that is expected, and the financial and societal burden that an increase in
AIDS will cause can be used to foresee possible futures and to test appropriate policy
responses.
1.2 Research Goals
The system dynamics model of the AIDS epidemic in Botswana and Uganda proposed here
will be used to look at the key economic and social indicators of labor force, the societal
burden resulting from a decreased workforce (burden per uninfected adult), the infected
population, AIDS spending per person, percentage of orphans, AIDS related deaths, and the
effect of AIDS on life expectancy to better visualize the effect that the AIDS has and will
have on the two countries. The effect of existing and new policies such as abstinence only,
educational campaigns, increased availability of ARTs and condoms, circumcision, and
family planning will be tested on the key indicators.
2 Methodology
2.1 Boundaries
Although there are many ways to contract the HIV virus, this study will focus on infections
resulting from sexual activity and mother to child transmission (MTCT). Also, the study only
focuses on Botswana and Uganda. The dynamics of contracting the HIV virus, and later
developing AIDS is very complex and should involve several separate stock flow structures
for different stages of the disease as well as different age structures. The complexity of the
HIV virus was simplified for this model and the effect of early detection and treatment of HIV
was left out. Since the year 2030 is a key year for Millennium Development Goals, the time
span of the model is from the approximate start of the AIDS epidemic in 1980 to 2030.
2.2 The model
In figure 3.2 on the next page, the complete structure of the model is shown as it was modeled
in Vensim. Some variables were given a background color, to indicate their function as a
policy tool (blue) or key performance indicator (yellow). There are some hidden layers in the
model that are not shown in Figure 2.1.
total money spent on ig —
extra years to live by
Figure 2.1
2.3. Main structure
As adapted from , the main structure of the HIV/AIDS epidemiology model consists of a
susceptible population stock (uninfected adults), from which people flow via infections to an
HIV-infected population stock (infected adults), and then through an AIDS manifestation flow
to aclinical AIDS population stock, after which they die (deaths by AIDS). Since the entire
population does not die from AIDS, in addition to the Dangerfield model (which is used for
HIV/AIDS in the UK), there is another outflow from uninfected adults, namely non-AIDS
deaths. Also, in contrast to the Dangerfield model, births, uninfected children and maturation
were taken into account. The main stock-flow structure can be seen in Figure 2.2.
HIV-infected adults can pass the HIV virus on to their newborn children during pregnancy,
labor and delivery, or breastfeeding (if the infant comes in contact with its mother’s body
fluids such as blood through a cut on the breast). The passing of the HIV virus from mothers
to children is called mother-to-child-transmission (MTCT) and the probability of MTCT is
estimated at between 20% and 50% . In the current model, the median value of 35% was
taken as the probability of MTCT. The most important implication of MTCT for the model
structure is that it requires an additional infected births flow into an infected children stock.
For simplicity, and because most of the infected children die before they mature, there was no
connection made between infected children and infected adults. It was assumed that the
infected children die (child deaths by AIDS) at an average age of life expectancy of child born
with HIV.
5 = uminfected | [uninfected | [infected] aac ae,
children _| saturation (—0ults J" ction -— AtDs ~ | population | deaths by L__AIDS
manifestation AIDS
infected
infected births L_tildven_| cpa deaths by
AIDS
Figure 2.2 Main stock-flow structure of the HIV/AIDS epidemiology model
This model is dealing with a timespan that is partially historical and partly in the future.
Variables of which the historical values are known (e.g. fertility rate) start as an exogenous
lookup function, and continue from 2012 on endogenous variables. Therefore, a lookup switch
was used to let the historic data overflow into the model-calculated data for the future. This
lookup switch function goes from 0 in 2010 to 1 in 2012. Every variable that uses historic data
was made up of a [variable name] history multiplied by (1-lookup switch), and a [variable
name] future multiplied by lookup switch.
Although the basic structure of the HIV/AIDS models are the same, the initial values and
lookup functions for history are different for Uganda and Botswana. Besides this difference,
the model works the same for both countries and could be modified for other countries with
high HIV prevalence.
2.4 Infections flow
The most sophisticated and perhaps most important formula used in the model is the one used
for the infections flow. For this formula we were inspired by the paper by ; the formula is
shown below.
IN essays a 3 Nacts (Alife Aconsent)
x Neaeeaors GAP andomBiondom ete AtitemAconsent)
I= Nuning Find FO Avmoomn = (1-1 Pirans) Npartners )
Atite — Aconseni
where:
I = infections
Aconsent = age of consent
Atite = life expectancy
Econdom = effectivity of condom use
Fondom = percentage condom use
Fins = HIV prevalence adults
Nacts = average number of sexual acts per year
Npartners = average number of sexual pariners in lifetime
unin uninfected adults
Ptrans = probability of HIV transmission
The average life expectancy of people in Botswana and Uganda lies between 50 and 60 years
and for simplicity the sexually active lifetime is taken as life expectancy — age of consent, the
latter of which a value of 17 years was used. Thus it was assumed that adults remain sexually
active until they die.
2.5 Factors influencing the infections rate
It was assumed that the probability of HIV transmission during heterosexual intercourse is
between 0.2% and 1.0% per act , depending on the degree of risk behavior. The risk behavior
multiplier is modeled as a linear function of the awareness of HIV (percentage aware of
having HIV). For Uganda, this percentage gradually increases from 0% in 1980 to 80% in
2012, following an asymptotic curve. For Botswana, the percentage aware of having HIV in
2012 was assumed to be 90%. Since this is a strong assumption, sensitivity testing on this
variable was performed (see section 4: Validation).
Circumcision lowers the probability of HIV transmission by 60% . A higher percentage
circumcised therefore serves as a policy tool to lower the infections rate. Another important
factor is percentage condom use, in the current model calculated by acceptance of condom
use times availability of condoms. The effectiveness of condom use was shown through
research to equal 80% . Finally, the average number of sexual partners in lifetime was set to
4.5 and the average number of sexual acts per year was calibrated to 100, to fit the actual
infections rate data.
2.6 Treatment of HIV
As of 2004, anti-retroviral treatment (ART) has become available to lengthen the incubation
time of HIV . In other words, ART postpones the manifestation of AIDS symptoms in HIV
patients. Before ART became available, the average incubation time of HIV was estimated at
8 years . The value of the variable extra years to live by using ART is uncertain and cannot be
proven since the treatment exists for only 9 years now. A value of 20 years was taken as a
starting point since it results in the closest fit of the model-calculated deaths by AIDS with the
numbers of deaths by AIDS reported by the .
With the introduction of ART, MTCT can also be prevented. An increase in prevention of
mother-to-child-transmission (PMTCT), e.g. a higher PMTCT coverage, leads to a lower rate
of infected births.
The World Bank provides data for the development of the percentage on ART and PMTCT
coverage . From 2012 on, different future scenarios can be tested by adjusting the values of
percentage on ART increase/year and PMTCT coverage increase/year. For the base run, a
status quo was simulated by setting these values to 0%.
2.7 Life expectancy
The historic data of life expectancy in Uganda and Botswana is known, and it can be observed
in earlier years that life expectancy declines when HIV prevalence increases. Since life
expectancy serves as an important performance indicator, there is a need to acquire values for
future life expectancies dependent on HIV prevalence. By reasoning, a function for life
expectancy can be obtained in which the total population is divided by the total number of
deaths per year:
total population
life expectancy future = +255) deaths per year
After comparing the course of this function to actual life expectancy data, it was observed that
this function follows the pattern of history closely. However, the value of this endogeously
calculated life expectancy was at every point higher than the actual life expectancy. After
calibration of life expectancy history with life expectancy future, a multiplier with a value of
0.73 was found for Uganda that corrects for the unwanted upward shift. The result can be seen
in Figure 2.3. The multiplier also worked for the Botswana data.
Life expectancy
60
55
S50
45
40
1980 1984 1988 1992 1996 2000 2004 2008
Time (Y ear)
life expectancy future : base run
life expectancy history : base run
Figure 2.3 Comparison of life history and model life
2.8 Abstinence
Abstinence is considered an important policy in fighting AIDS, especially in Uganda. Y oung
people are promoted to practice abstinence, i.e. to save sex for marriage and to stay with only
one partner during their lifetime. Although “abstinence only” is controversial among the
development and international community, it is strongly promoted by the Ugandan
government. Therefore, a structure was added to the model in which a variable percentage of
children flow to an abstinent population after maturing (see Figure 2.4). The assumption was
made that people in this abstinent population will never get HIV. Initially, the percentage
practicing abstinence was set to 0%.
Figure 2.4 Structure for abstinent population
2.9 Financial burden of HIV/AIDS
All measures to fight the HIV/AIDS epidemic cost significant money either from
governmental or private sources. This includes money for ART and PMTCT, and money for
making condoms more available. Numbers for the costs of these measures are estimated based
on different sources or guesses, to examine how different policy measures increase financial
burden. Since the costs are estimates a sensitivity analysis was performed to examine how
changes in the estimations affect model behavior (see section 4: model behavior).
2.10 Burden per uninfected adult
The burden per uninfected adult is a performance indicator calculated by:
infected adults + infected children + uninfected children + clinical AIDS population
uninfected adults + abstinent population
In short, it divides the help-demanding population by the help-giving population.
3 - Validation
3.1 Comparison with real data
The most logical first validation test would be a comparison between the numbers obtained by
running the model and the numbers found in databases . As a first test, the reported HIV
prevalence in Uganda was compared (see Figure 3.1) and corresponds well with existing data.
0.05
1980 1984 1988 1992 1996 2000 2004 2008 2012
Time (Y ear)
reported HIV prevalence model : base run
reported HIV prevalence W orldbank : base run
Figure 3.1 Comparison of model-calculated HIV prevalence and actual HIV prevalence in Uganda
Variables were compared with actual data, for both Uganda and Botswana, and the overall fit
between the model and the available statistics was sufficient for exploratory and policy
analysis within the model.
3.2 Extreme value testing
Since the effect of changes to the system on future outcomes is important, extreme value
testing was done to check for unrealistic behavior. The Botswana model was used for an
extreme value test in which all policies are suddenly ceased in 2010, by setting the availability
of condoms, ARTs, and PMTCT to zero. As expected, the model-calculated infections go up
drastically in response to this cease of policies (Figure 3.2).
infections
60,000
45,000
§
: 30,000
15,000
0
1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030
‘Time (Y ear)
infections : cease all policies
infections : base nn
Figure 3.2 New infections per year in Botswana after ceasing all HIV policies
3.3 Sensitivity analysis
One of the most uncertain parameters in the model is percentage aware of having HIV;
therefore a sensitivity analysis was performed, using the Uganda model. Since it is a lookup
function, a direct Monte Carlo simulation is not possible in Vensim. Instead, five different
scenarios were tested. For every scenario, a different course of the graph for percentage
aware of having HIV was chosen. The different lookup functions used are plotted in Figure
3.3,
percentage aware of having HIV
1
0.75
05 =
0.25 <
-.
0
1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030
Time (Y ear)
percentage aware of having HIV : 120% of base un
percentage aware of having HIV : base nun
percentage aware of having HIV : 80% of base nn
percentage aware of having HIV : 60% of base nn
percentage aware of having HIV : 40% of base mn
Figure 3.3 Five courses of percentage aware of having HIV for five different scenarios
The effects of these different scenarios on Ugandan society were checked by looking at the
important performance indicator infections (Figure 3.4). Obviously, different assumptions
about the uncertain parameter of percentage aware of having HIV (and especially its value
after 2012) can have large consequences for the future behavior of the model. Although the
infections flow displays similar behavior for all tested values (i.e. it shows no behavioral
sensitivity), different values result in different numerical values for infections per year (it
shows numerical sensitivity).
infections
600,000
450,000
300,000
person/Y ear
150,000
0
1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030
‘Time (Y ear)
infections : 120% of base run
infections : base run.
infections : 80% of base rn.
infections : 60% of base run
infections : 40% of base rn.
Figure 3.4 Infections flow for five different scenarios
A check for the plausibility of these scenarios can be done by looking at the graph of reported
HIV prevalence (Figure 3.5), which is also a function of the percentage aware of having HIV
(since people who are unaware of their HIV status may not be reported). In this graph, the
historic data of reported HIV prevalence is added. The base run scenario (red line) matches
most closely with the historical data (light blue line). The closeness of fit can be a slight
justification for the use of the base run values for percentage aware of having HIV as
described in the previous section on specification.
1980 1984 1988 1992 1996 2000 2004 2008 2012
Time (Y ear)
reported HIV prevalence : 120% of base mun
potted HIV prevalence : base run
potted HIV prevalence : 80% of base run
reported HIV prevalence : 60% of base run
reported HIV prevalence : 40% of base run
reported HIV prevalence W orldbank : base run
Figure 3.5 Reported HIV prevalence for five different scenarios, compared to real reported HIV prevalence as
provided by WorldBank, 2012)
10
4 = Model behavior
4.1 Infections and HIV prevalence
A Latin Hypercube uncertainty analysis is performed to check the influence of variations in
values for uncertain parameters that are used to calculate the infections flow. These values
also indirectly affect the evolution of HIV prevalence. A range of values (see Table 4-1) was
chosen for the variables acceptance of condom use, availability of condom use, percentage of
males circumcised, average number of sexual acts per year, and average number of sexual
partners in lifetime, according to a random uniform distribution. The results are shown in
figures 4.1-4.4. As shown by the graphs, the future AIDS epidemic in Uganda may
dramatically increase regardless of values chosen.
Table 4-1 Parameter ranges used in sensitivity analysis
Parameter Uganda Botswana
Acceptance of condoms 30% — 50% 70% — 100%
Availability of condoms 30% — 50% 80% — 100%
Percentage males circumcised 0% — 20% 0% — 20%
Average no. of sexual acts per year 50-150 50 - 150
Average no. of partners 4-5 4-5
Uganda Botswana
base run base run
50% 75% I 95% FI 100% 50% 75% RN 95% FI 100%
infections infections
600,000 40,000
450,000 30,000
300,000 20,000
150,000 = 10,000
0 0
1980 1993 2005 2018 2030 1980 1993 2005 2018 2030
Time (Y ear) Time (Y ear)
Figure 4.1 Sensitivity graph for infections in Uganda Figure 4.2 itivity graph for ii ions in
11
base nn
5
base run
50% 75% NN) 95% [I 100% ae
0% 75% (IR 95% I 100%
HIV prevalence adults HIV prevalence adults
0.2 0.4
0.15 0.3
0.1 2 0.2
0.05 0.1
ogo 1993 2005 2018 2030 oso 1993 2005 2018
Time (Y ear) Time (Y ear)
2030
Figure 4.3 Sensitivity graph for HIV prevalence in Uganda Figure 4.4 Sensitivity graph for HIV prevalence in
4.2 Financial burden of HIV/AIDS in Botswana
Botswana
The exact costs of HIV/AIDS prevention and treatment practices are not known, so an
uncertainty analysis was performed on the values determining these costs, using the ranges
depicted in Table 4-2. The sensitivity graph for Botswana is shown in Figure 4.5. The
estimated spending per capita lies between 75 and 125 US dollars per year, and will increase
slowly in the next 20 years if there is no major change in HIV prevalence. The average values
for the costs of HIV/AIDS prevention and treatment practices were used in the rest of the
analysis.
base run
50% 75% 95% I 100%
"AIDS/HIV spending/capita'
200
150
100
50 |
0
1980 1993 2005 2018 2030
Time (Y ear)
Figure 4.5 Sensitivity graph for AIDS/HIV spending per capita (US dollars per year) in Botswana
Table 4-2 Parameter ranges used in sensitivity analysis
Costs in US dollar per person per year
Condoms 3-15
ART 500 — 1000
PMTCT 1500 — 2500
12
Since the availability of ART, PMTCT and condoms is already close to 100% in Botswana,
the future spending on HIV/AIDS is highly dependent on the HIV prevalence in the future;
therefore sensitivity analysis was also done on HIV/AIDS spending/capita in Botswana, using
the same range of values on the same parameters as for the infections and HIV prevalence
sensitivity graphs (Figure 4.6).
base run
50% 75% IN 95% FI 100%
"AIDS/HIV spending/capita"
200
150
100
oso 1993 2005 2018 2030
Time (Y ear)
Figure 4.6 Sensitivity graph of AIDS/HIV spending per capita (US dollar per year) in Botswana
4.3 Infected children
The uncertainty analysis for infected children shows that without policy measures the number
of infected children in Uganda will increase (Figure 4.7).
base run base run
50% 75% (RN) 95% I 100% 50% 75% (A 95% FI 100%
infected children infected children
800,000 40,000 |
600,000 30,000
400,000 20,000
200,000 10,000
os 1993 2008 2018 2030 © ©» 19801993 2005 2018 2080
Time (Y ear) Time (Y ear)
Figure 4.7 Sensitivity graph for infected children in Figure 4.8 Sensitivity graph for infected children in
Uganda Botswana
13
5 __ Policy Analysis
In the past and present, both Uganda and Botswana have been hailed as success stories in
their reduction of HIV prevalence. In the early 2000s Botswana aggressively attacked the
AIDS problem through a combination of international and governmental funding, resulting in
a decrease of new infections and HIV prevalence. Both Botswana and Uganda have promoted
the policy of ABC—Abstinence, Be Faithful, and Condom Use. In recent years Uganda has
focused mainly on the abstinence portion of this policy, a move that has drawn much criticism
from the international community (Avert, 2012). Some studies speculate that the focus on
abstinence only has resulted in Uganda’s increase in total infected population and constant
HIV prevalence rate over the last ten years.
Another new area in AIDS research is the effect of circumcision on the spread of AIDS. New
studies have shown that circumcision can reduce the chances of contracting HIV by
approximately 60% (WHO, 2012). The effect of circumcision on the epidemiology of AIDS
was tested here to evaluate its effectiveness as an AIDS reduction policy.
Lastly, three other crucial policy responses are deployment of ART, the prevention of mother
to child transmission, and reduction of fertility rate (more applicable for Uganda). Full
availability of ART and Prevention of MTCT have almost been achieved in Botswana, but in
Uganda treatment is lacking; therefore one appropriate policy response would be to scale up
ART and PMTCT services. Also for Uganda, creating national family planning policies to
reduce the astronomical population growth has potential to decrease the HIV prevalence.
Currently, contraceptive use and availability is extremely low (WHO, 2012). Unfortunately,
due to the non-ideal way of modeling delays in population flows, and the lack of a distinction
between different age groups in the adult population, the current model is not suited to test
this policy appropriately.
Table 5-1 Values used for policy variables
Values used in policy analysis Uganda Botswana
Abstinent population 30% 30%
Acceptance of condoms 90% 100%
Availability of condoms 90% 100%
Percentage males circumcised 80% 80%
Average no. of partners 3 3
ART increase NA 5%
PMTCT increase NA 5%
To test policies, realistic values were chosen for the various policy variables. Since abstinence
only is a highly controversial policy, the percentage of those practicing abstinence was set to
30%. As shown in the following figures abstinence has a positive impact on reducing
infections, and a low cost per person. The number of partners was reduced from 4.5 to 3 to
test the policy of promoting faithfulness. Condom use was tested with different values for
Botswana and Uganda since Botswana already has high availability and acceptance.
Acceptance and availability were set to 100% for Botswana and 90% for Uganda (a big
14
increase from the current value of 40%). For both countries the percentage of the male
population circumcised was set to 80%. Lastly, for Uganda, ART and PMTCT coverage was
increased 5% per year. The effect of these policies on the key performance indicators is
shown in the following graphs. Table 5-1 summarizes the policy values used in the policy
analysis.
5.1 ABC Policies on KPIs Botswana
Infections & HIV Prevalence
Figure 5.1 shows how Botswana’s new infections have mostly leveled out regardless of
policies, but in order to achieve the nearly impossible target of no new infections in 2016,
Botswana will need to scale up their existing policies. Figure 5.2 shows that Botswana’s HIV
prevalence has mostly leveled out unless all existing policies are combined and scaled up.
infections
et
1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030
Time (Year)
0%
infections : be faithful: avg 3 partners
infections : condom use 100%
HIV prevalence adults
1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030
HIV
HIV abstinence 30%
HIV be faithful: avg 3partners
HIV prevalence adults : condom use 100%
HIV
fections : he
HIV basemun
Figure 5.1 Effect of policies on new infections per year in
Botswana
Figure 5.2 Effect of policies on HIV prevalence in Botswana
Burden per Uninfected Adult
The burden per uninfected adult shows a decrease in the future, likely due to the positive
population growth in both countries. However, the curve follows a similar pattern to the HIV
prevalence and number of infections.
burden per uninfected adult
2
1.65
B13
0.95
0.6
1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030
Time (Y ear)
burden peruninfected adult: all policies combined
burden per uninfected adult : abstinence 30%
burden per uninfected adult : be faithful: avg 3 partners
burden peruninfected adult :condomuse 100%
burden peruninfected adult
burden peruninfected adult : base run
Figure 5.3 Effect of policies on burden per uninfected adult in Botswana
15
AIDS spending per capita
Since Botswana already spends a large portion of government money on AIDS prevention
Figure 5.4 shows that there isn’t a large difference in cost per person for the different policies
and overall spending will go down per person.
AIDS/HIV spending/capita
200
dollay/(person*Y ear)
S
S
1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030
Time (Y ear)
"AIDS/H
“AIDS/HIV spending/capit
“AIDS/HIV spending/capit
“AIDS/HIV :
“AIDS/HIV g :base run
Figure 5.4 Effect of policies on AIDS/HIV spending/capita in Botswana
“AIDS/HIV spending/capita"
HIV
5.2. ABC Policies on KPIs in Uganda
Infections & HIV prevalence
Unlike Botswana, there is significant room for policy improvements in Uganda’s minimal
existing HIV prevention policies. Figure 5.5 shows that unless Uganda starts creating
aggressive AIDS reduction policies an increase in new HIV infections could be expected.
infections
600,000
450,000
4
g
$300,000
E as
150,000 |
0
1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030
Time (¥ ear)
infections
infections
infections : abst
infections : be faithful: avg 3 partners
fections :circumision 80%
infections : ART 5% ps
infections : PMTCT 5% increase per year
infections : base run
Figure 5.5 Effect of policies on new infections per year in Uganda
16
Similar to the results from new infections, Figure 5.6 shows that if Uganda fails to update
their HIV policies the HIV prevalence will increase.
HIV prevalence adults
Dani
0
1980 1985 1990-1995 2000-2005 = 20102015 +2020 2025 += 2030
ime (Y ear)
HIV
HIV
‘ABC combined
stinence 30%
HIV
HIV prevalence adults : be faithful: avg 3 partners
HIV
HIV
HIV
ART
HIV PMTCT 5% Pl
HIV prevalence adults: base run
Figure 5.6 Effect of policies on HIV prevalence Uganda
Burden per uninfected adult
The burden per uninfected adult in Uganda also shows (Figure 5.7) a slow decrease in the
future, probably due to Uganda’s continually growing population. However, without policy
measures the decrease is less severe and mostly levels out.
burden per uninfected adult
1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030
burden per dult: all policies combined
burden per uninfected adult :ABC combined
burden per ult : abst
burden per uninfected adult : be faithful: avg 3 partners
burden per uninfected adult : condom use: 90%
burden per uninfected adult : circumcision 80%
burden p PART 5 pery’
burden per uninfected adult : PMTCT 5% increase per year. — $$$
burden per adult : base rin
Figure 5.7 Effect of policies on burden per uninfected adult in Uganda
17
AIDS spending per capita
Conversely to Botswana, Figure 5.8 demonstrates that AIDS spending in Uganda will
increase (again could be attributed to the population increase). In particular, increasing
coverage of ARTs (an expensive treatment option) will greatly increase spending. The need
for more AIDS spending in Uganda predicted by experts is matched by the behavior of this
model.
AIDS/HIV spending/capita
60
45
i
4% 30
&
z
3
15
0
1980 1985 1990 1995 2000 2005 2010 2015 2020 +2025 ~~—2030
Time (Yeat)
“AIDS/HIV: all poli bined
“AIDS/HIV ABC combined
’ etnence 90%
i
i
DS/HIV
‘AIDS/HIV spending/capita"
“AIDS/HIV
be faithful: avg 3 partners
cine 80%
“AIDS/HIV spending/capita"
“AIDS/HIV spending/capita"
ART 5% increase peryear
PMTCT 5% increase peryear
“AIDS/HIV base run
Figure 5.8 Effect of policies on AIDS/HIV spending per capita (US dollar per year) in
Uganda
Deaths by AIDS
Since Uganda does not have universal coverage of ARTs it is interesting to examine how the
deaths by AIDS flow changes given changes in policies. Like the previous graphs Figure 5.9
shows how the most effective policy is a combination of AIDS prevention measures (blue
line).
deaths by AIDS
200,000
150,000
g
= 100,000
B aa
& ——
++
50,000 |
1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030
Time (Year)
deaths by
deaths by Al bined
deaths by "6
deaths by AIDS: be faihfu- avg 3 partners
deaths by
y AIDS :ART 5
deaths by AIDS: PMTCT 5%
deaths by AIDS :b
Figure 5.9 Effect of policies on deaths by AIDS in Uganda
18
Infected Children
Uganda has very low PMTCT coverage, therefore the effects of increasing PMTCT or simply
increasing AIDS prevention has large impacts on the growth of infected children (shown in
Figure 5.10).
infected children
400,000
300,000
# 200,000
100,000
0
1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030
Figure 5.10 Effect of policies on number of infected children in Uganda
5.3 Policy Conclusions
As expected, the AIDS epidemic in Botswana has leveled out and even with increased policy
action the, positive effects on the epidemic will not be as pronounced. Maximizing existing
policies and introducing new policies such as circumcision will assist Botswana in reaching
the unlikely goal of no new infections by 2016.
As exhibited consistently in the model behavior, the AIDS epidemic in Uganda is going to
increase and will only slow given strong policy action. In a more positive light, since Uganda
is lacking strong and effective prevention policies there is significant room for expansion in
the AIDS fight. If Uganda combines all existing policies and introduces circumcision, the
HIV prevalence will decrease and other negative effects of the epidemic will be less
pronounced.
6 Conclusions and Discussion
6.1 Conclusions & Recommendations
Based on the model behavior several conclusions can be made regarding the dynamics of the
future AIDS epidemic and appropriate policy responses. It seems likely that the HIV
prevalence and number of infections in Uganda will increase. Although the popular policy in
Uganda of abstinence shows some improvements in the key performance indicators, it is
difficult to appropriately test since it depends on the assumption of the percentage of people
who are actually practicing abstinence. The most effective policy in both countries is a
combination of scaling up PMTCT, condom use and availability, circumcision, and ART
coverage. For Botswana the effects of scaling up HIV prevention are not as extreme since
universal coverage of PMTCT and ARTs have already been reached. It might be useful for
Botswana to invest more in new prevention methods such as circumcision on top of existing
policies. In Uganda there is much room and need for increased development in addressing the
19
AIDS epidemic. Universal coverage of ARTs and PMTCT has potential to slow the AIDS
epidemic in Uganda but currently coverage is severely lacking. Additionally, as Uganda’s
population increases, redirecting HIV prevention away from simply abstinence only will be
essential in order to address the potential increase in infections and HIV prevalence.
Lastly, all policy efforts to fight AIDS come at a cost. Botswana already invests significant
funds in HIV prevention but the difficulty in the future will be balancing the social cost that
the AIDS epidemic presents with the economic burden it places on a country that seeks to
slow the reach of the virus. Furthermore as international aid will possibly decrease in the
coming years with the advent of the global financial crisis, as a result, both Uganda and
Botswana will feel an increased financial burden due to AIDS. In order to stop the inevitable
spread of the virus Uganda will certainly need to expand HIV treatment and prevention and
Botswana will need to begin reevaluating existing policies for both their effectiveness and
affordability, rather than solely letting the government absorb the cost of AIDS.
Ultimately, the devastating effect of AIDS on the social and economic development of
Botswana and Uganda creates the need for both countries to radically attack the spread of the
virus through concerted policy efforts. In addition to simply improving implementation of
existing and new HIV prevention policies, focusing on reducing the fertility rate in Uganda
could have a positive impact on the development of HIV. Finally, although Botswana has
been praised for their success in reducing HIV, future improvements to meet their set targets
will be difficult and increasingly costly to the government.
6.2 Future research
In the current model, different methods were used to model outflows from the population
stocks. In some cases, a first-order delay was used, while in other cases the value of the stock
was divided by the average time after people leave the stock. The inconsistency in modeling
outflows was necessary for the current model to function, but is not ideal.
Another limitation of the current model is that there is only one stock for adults, meaning the
fact that partners are usually about the same age is not taken into account. It is assumed that
17 year olds will have sex with 50 year olds, as frequently as 24 year olds will have sex with
24 year olds. In a more sophisticated model, a split up of the adult population into different
age groups is recommended. The two limitations of the current model explain why the effects
of family planning policies (as discussed in chapter 6) cannot be tested reliably. A fix of both
limitations in a future study would solve these problems and broaden the possibilities for
policy testing.
Furthermore, the HIV/AIDS epidemic has huge effects on the labor force and ultimately
productivity of a country. Using the Cobb-Douglas production function as a guide to examine
how changes in labor force and productivity affect GDP growth, future work on the dynamic
of GDP/capita with respect to AIDS should be done. Since both Botswana and Uganda have
high growth rates as their economies are still growing, it would be valuable to explore how
AIDS will affect their future economic growth.
Another important aspect of AIDS prevention and treatment that was overlooked in this
model is the timely testing and follow-up treatment of potential HIV positive people (Highly
Active Antiretroviral Therapy-HAART). In Uganda, availability of testing services is
extremely low and an increase in testing could have a positive impact on reducing the HIV
prevalence (or conversely cause an increase in the reported HIV prevalence). For future
20
models adding an additional stock-flow structure to model HIV testing and HAART treatment
would provide a more thorough policy analysis of the AIDS epidemic.
Much of HIV research has focused on prevention and treatment but steered away from the
possibility of a cure for AIDS. It could be beneficial to examine the social and economic
effects of curing AIDS in Uganda and Botswana since a cure for AIDS could not only
positive but negative effects (with regards to un-damped population growth) on the future
development of both countries.
Finally, since many of the values and assumptions used in this analysis are highly uncertain
and should be tested under deep uncertainty, future work will be done using the Exploratory
Modelling and Analysis (EMA) tool developed by E. Pruyt and J.H. Kwakkel.
7 References
Avert. (2012). The HIV and AIDS epidemic in Africa. Retrieved from http://avert.org/A IDS-
hiv-africa.htm
Auvert, Bertran, Dirk Taljaard, Emmanuel Lagarde, Joélle Sobngwi-Tambekou, Rémi Sitta,
and Adrian Puren. "Randomized, Controlled Intervention Trial of Male Circumcision
for Reduction of Hiv Infection Risk: The Anrs 1265 Trial." PLoS Med 2, no. 11
(2005): e298.
Boily, Marie-Claude, Rebecca F. Baggaley, Lei Wang, Benoit Masse, Richard G. White,
Richard J. Hayes, and Michel Alary. "Heterosexual Risk of Hiv-1 Infection Per Sexual
Act: Systematic Review and Meta-Analysis of Observational Studies." The Lancet
Infectious Diseases 9, no. 2 (2009): 118-29.
Dangerfield, Brian, and Carole Roberts. "A Role for System Dynamics in Modelling the
Spread of Aids." Transactions of the Institute of Measurement and Control 11, no. 4
(November 1, 1989 1989): 187-95.
De Cock, K.M., M. Fowler, E. Mercier, I. De Vincenzi, J. Saba, E Hoff, D.J. Alnwick, M.
Rogers, and N. Shaffer. "Prevention of Mother-to-Child Hiv Transmission in
Resource-Poor Countries: Translating Research into Policy and Practice." JAMA: The
Journal of the American Medical Association 283, no. 9 (2000): 1175-82.
Johnson, Leigh. "An Introduction to the Mathematics of Hiv/Aids Modelling." In, (2004).
http://www.commerce.uct.ac.za/Research_Units/CARE/RESEA RCH/PA PERS/Maths
Intro.pdf.
Lule, Elizabeth, and Markus Haacker. The Fiscal Dimension of Hiv/Aids in Botswana, South
Africa, Swaziland, and Uganda. Washington, D.C.: The World Bank, 2012.
Weller, S. C., and K. Davis-Beaty. "Condom Effectiveness in Reducing Heterosexual Hiv
Transmission." Cochrane Database of Systematic Reviews 2007, no. 4 (2007).
WorldBank. "World Databank: World Development Indicators (Wdi) & Global Development
Finance (Gdf)." The World Bank Group, 2012.
21
