PANDEMICS: AIDS

A. Understanding HIV and AIDS: An introduction for actuaries

This section was written by LF Johnson (updated May 2023).

1. Why is it important for actuaries to know about HIV/AIDS?

In South Africa, AIDS currently accounts for around 50 000 deaths per annum (roughly
8% of all deaths). Actuaries working in South Africa need to have a clear understanding
of how this disease is affecting mortality rates and disability rates, and which sections
of the population are most severely affected by the disease. In order to manage the
financial risks associated with HIV/AIDS, it is also important that actuaries have a good
understanding of the strategies for preventing and treating HIV infection.

2. What is HIV/AIDS?

The human immunodeficiency virus (HIV) is a virus that destroys human immune cells,
thus bringing about a severe weakening of the immune system. In Africa, the virus is
mainly transmitted through sexual contact, but it can also be transmitted through
injections with contaminated needles, through blood transfusion and in the process of
childbirth and breastfeeding.

HIV is genetically very similar to immunodeficiency viruses that are found in

chimpanzees and monkeys, and the most likely explanation for how HIV originated is
that it is the result of multiple transfers of these immunodeficiency viruses from
primates to humans. There are two types of HIV: HIV-1 and HIV-2. HIV-2 occurs
almost exclusively in West Africa and is very rare in other parts of the world.
Individuals who are infected with HIV-2 survive for longer than those infected with
HIV-1, and it would also appear that people infected with HIV-2 are less likely to
transmit the virus than individuals infected with HIV-1. HIV-1 accounts for the vast
majority of the global HIV/AIDS burden, and most of the discussion that follows relates
specifically to HIV-1.

The acquired immune deficiency syndrome (AIDS) is a collection of illnesses that

occur in the late stages of HIV infection, when the patient’s immune system has been
severely weakened by the virus. Common AIDS-defining illnesses include
extrapulmonary tuberculosis, pneumonia and wasting syndrome. In the absence of
treatment, most individuals die within a year or two of the first AIDS-defining illness.

215
3. What factors affect the risk of HIV infection?

A host of factors affect the individual’s risk of HIV infection. These factors can be
divided into three classes: biological factors, sexual behaviour factors and socio-
economic factors.

(a) Biological factors

One of the most significant factors affecting the risk of acquiring HIV is infection with
other sexually transmitted diseases (STDs). These STDs cause symptoms such as ulcers
and lesions, which provide a more direct portal of entry for HIV. STDs also cause an
increased concentration of immune cells in the genital tract, and HIV requires these
immune cells in order to replicate and establish itself. STDs therefore increase
susceptibility to HIV substantially. HIV-infected individuals who have other STDs also
have higher concentrations of HIV in their genital tract, which makes them more likely
to transmit the virus. Since a high proportion of STDs go unrecognized and untreated,
they contribute significantly to the spread of HIV, particularly in countries where access
to high-quality STD treatment is poor.

Women are generally at a higher risk of HIV transmission, per act of sex, than men.
This is largely due to the surface area of the female reproductive tract being larger than
that of the male reproductive tract, but it is also partly due to gender inequality
(discussed below). There is some evidence to suggest that young women may be at a
particularly high risk of HIV infection due to their reproductive tracts being less mature,
but this has not been proven conclusively.

Men who are circumcised are at a significantly lower risk of acquiring HIV and other
STDs than men who are uncircumcised, and studies conducted in Africa have tended
to show that in societies in which male circumcision is traditionally practised, the level
of HIV prevalence is lower than in societies in which male circumcision is not practised.

Genetic factors can also influence susceptibility to HIV, and a small proportion of
people are actually naturally resistant to HIV because their immune cells do not have
the receptors that HIV requires in order to attach itself to these cells.

(b) Sexual behaviour factors

Gender inequality is a significant factor driving the spread of HIV in many countries.
Because of limited employment opportunities for women in many developing
countries, many women are forced to rely on transactional sex (sex in exchange for
food, gifts or money) as a basic survival strategy. Women are often financially
dependent on male partners, and in many cases their partners are several years older
than themselves. These gender power imbalances make it difficult for women to insist
on condom usage or to confront their partners about suspected infidelities. Studies
conducted in Africa show that many women who know they are HIV-positive do not
inform their spouses because they are afraid that their partners will react violently or
abandon them and leave them without financial support. Many women experience
violence in sexual relationships, and the high incidence of rape in many countries is a
further factor increasing women’s susceptibility to HIV.

216
It has been hypothesized that a significant factor promoting the spread of HIV is high
levels of partner concurrency (having more than one partner at the same time). Because
individuals infected with HIV are most infectious during the first few weeks of
infection, a person with two partners who acquires HIV from one partner is likely to
transmit the virus to their other partner within a short space of time – and if that partner
has another partner, they too are likely to transmit the virus very rapidly. The virus
therefore spreads much more rapidly in a population in which there are high levels of
partner concurrency than in a population in which serial monogamy is the norm, even
if the populations are similar in terms of the average number of partners people have
per annum.

(c) Socio-economic factors

In the early stages of the HIV epidemic, a large number of African studies found that
HIV prevalence levels were higher in more educated individuals than in less educated
individuals. This may be explained by the fact that people of higher socio-economic
status find it easier to attract partners, which places them at a greater risk of acquiring
HIV. However, more recent studies suggest that the extent to which higher educational
attainment increases the risk of HIV diminishes over time, and recent studies have even
found that the risk of HIV is higher in less educated individuals than in more highly
educated individuals. This could be because there have been relatively more HIV/AIDS
awareness programmes in recent years, and more highly educated individuals are more
likely to be exposed to these programmes. Individuals of higher socio-economic status
are also more capable of adopting measures to protect themselves against HIV (e.g.
consistent use of condoms).

In Africa, HIV prevalence tends to be higher in urban areas than in rural areas. This is
partly due to the greater opportunities for sexual networking in urban areas, but may
also be due to the fact that the epidemic has typically started in urban areas and
gradually filtered through to rural areas. In South Africa, HIV prevalence rates in rural
areas were initially lower than those in urban areas, but have since caught up with urban
prevalence rates, due to the high levels of migration between urban and rural areas.

4. How does HIV infection progress and how long do people survive with HIV?

Following initial HIV infection, an individual may experience glandular fever-like

symptoms that last for a few weeks. During this time, the so-called ‘window period’,
an individual will test negative for HIV on antibody tests. It is only after the individual
has seroconverted (i.e. started to produce antibodies to the virus), typically 3 to 4 weeks
after the initial infection, that these tests will yield positive results. Following the
passing of these initial symptoms, the individual enters a prolonged asymptomatic
phase, which typically lasts 4 to 6 years. In the absence of treatment, the individual then
starts to experience intermittent symptoms such as weight loss, diarrhoea and oral
infections. Finally, when the individual’s immune system has been severely weakened
by the HIV infection, they experience a variety of opportunistic infections (OIs), such
as extrapulmonary tuberculosis, which are regarded as being defining of AIDS.

The median time from infection to death is usually estimated to be between 9 and 11
years, in the absence of antiretroviral treatment. Individuals acquiring HIV at younger

217
ages tend to survive for longer with HIV infection than those acquiring HIV at older
ages. Genetic factors and nutritional factors also affect the length of time individuals
survive with HIV.

A number of laboratory tests have been used to determine the prognosis of people
infected with HIV. The two laboratory measures that are most commonly used are the
viral load and the CD4+ lymphocyte count. The CD4+ count is a measure of the
strength of the immune system; an uninfected individual would typically have a CD4+
count above 800 cells per mm3, while an individual experiencing AIDS would usually
have a CD4+ count below 200 cells per mm3. The viral load is a measure of the
concentration of HIV in the body, and can be thought of as determining the rate of
decline in the CD4+ count. The figure below shows the typical changes in viral load
and CD4+ count over the course of HIV infection, in the absence of treatment. Viral
load levels (often measured on a log scale) tend to be high at the time of seroconversion,
and then fall after the individual’s immune system starts to produce antibodies and other
immune cells to attack the virus. The viral load is important not only as a prognostic
marker, but also as a measure of an individual’s infectiousness; individuals with high
viral load levels are most likely to transmit HIV.

7 1200

CD4+ T lymphocyte count (cells/mm3)

Viral load
Viral load (log10 of # RNA copies/ml)

6
CD4 count 1000

5
800
4
600
3
400
2 Viral set
point
1 200

0 0
0 2 4 6 8 1 3 5 7 9 11
Weeks Years

HIV-infected individuals are often categorized according to their ‘WHO clinical stage’.
This categorization, developed by the World Health Organization (WHO), places HIV-
positive individuals into one of four disease stages, depending on the severity of the
symptoms that they have experienced. Stages 1 and 2 are defined in terms of minor
symptoms such as swollen glands and skin rashes, while stage 3 is defined in terms of
symptoms such as weight loss, oral infections and diarrhoea. WHO stage 4 is equivalent
to AIDS. Individuals are always classified in terms of the most severe symptoms they
have experienced up to the current time, and thus (for example) an individual cannot
go from stage 3 to stage 1, even if their stage 3-defining symptoms disappear
completely.

Although individuals infected at young ages tend to survive for longer than individuals
infected at older ages, children who are infected at or before birth are an exception.

218
These children experience very high AIDS mortality during the first year of life, as their
immune systems do not have a chance to become established before the virus attacks.
Children infected through breast-milk survive for longer than those infected at or before
birth.

5. How effective is treatment for HIV?

Currently there is no easy cure for HIV/AIDS (a few of people have been cured after
having received bone marrow transplants from people with genetic resistance to HIV,
but this operation has significant mortality risks). However, antiretroviral drugs can
limit the replication of the virus and thus bring about a reduction in the HIV viral load
and a restoration of the immune system. Early antiretroviral regimens consisted of only
one or two drugs (monotherapy and dual therapy respectively), but since the
development of new classes of drugs in the mid-1990s, more effective regimens of three
or more drugs (highly active antiretroviral treatment, or HAART) have been introduced.
These HAART regimens have been shown to reduce AIDS mortality rates to very low
levels; people who start HAART in the early stages of HIV infection may have a life
expectancy close to that in the general population. Individuals who only start HAART
in the extremely late stages of infection, however, are less likely to benefit from the
drugs, as it typically takes some time for the immune system to reconstitute itself, and
if the individual is already very sick, this reconstitution may not be fast enough to save
the individual from death.

Other forms of treatment include prophylaxis (treatment to prevent the onset of

common opportunistic infections) and micronutrient supplementation. Although these
forms of treatment are important, they generally have not reduced AIDS mortality rates
by more than 30%. It is therefore important that these other forms of treatment not be
viewed as alternatives to HAART.

6. What strategies are there for preventing HIV?

Most obviously, HIV prevention programmes need to begin with information and
education campaigns to build awareness around HIV and its consequences, and to
correct the many misconceptions regarding the transmission of HIV and the course of
HIV infection. Social marketing programmes promote the adoption of less risky sexual
behaviours, through the promotion and distribution of condoms, through mass media
campaigns and through school education programmes.

Voluntary counselling and testing (VCT) is a service provided to individuals wishing

to learn their HIV status. It includes pre- and post-test counselling, and has been shown
to lead to significant changes in sexual behaviour in individuals with positive test
results.

Improved treatment of STDs is also an important strategy for limiting the spread of
HIV, due to the role of other STDs in the transmission of HIV. In addition to improving
access to and quality of treatment, it is necessary to educate the public regarding the
recognition of STD symptoms and the importance of seeking prompt treatment for these
symptoms. A major problem is that many STD cases are asymptomatic (i.e. the

219
individual does not experience any symptoms suggestive of an STD), and these STDs
therefore go untreated for long periods. Another significant problem is that many
individuals do not inform their partners when they are diagnosed with an STD, with the
result that they are at a high risk of becoming reinfected after treatment.

Prevention of mother-to-child transmission (PMTCT) is a strategy to reduce the risk of

mother-to-child transmission of HIV, through the provision of antiretroviral drugs to
the mother before and after birth. In the absence of PMTCT, roughly one third of
children born to HIV-positive mothers are HIV-infected, either around the time of birth,
or through breast-feeding. PMTCT programmes can theoretically reduce transmission
rates to 2% or less if all pregnant women start long-term HAART prior to delivery.

Although HAART has for a long time been seen as a treatment strategy rather than a
prevention strategy, there has been growing interest in the use of HAART in prevention.
This is because HAART reduces the concentration of HIV in the body to very low
levels, and thereby reduces the infectiousness of HIV-positive individuals. A landmark
trial in serodiscordant couples (couples in which one partner is HIV-positive and the
other is HIV-negative) showed that HIV-positive individuals who were receiving
HAART were 96% less likely to transmit HIV to their sexual partners than HIV-
positive individuals who were not receiving HAART.

Antiretroviral drugs have also recently been shown to be effective in prevention when
used by HIV-negative individuals. Pre-exposure prophylaxis (PrEP) refers to the use
of antiretroviral drugs prior to HIV exposure to prevent HIV acquisition. Truvada, the
drug that is most commonly used in PrEP, has been available in South Africa for several
years. Truvada is highly effective in preventing HIV if it is used consistently, and has
few side effects. More recently, injectable PrEP (taken every two months rather than
daily) has been shown to be more effective than Truvada, but this is not yet available
in South Africa.

Little progress has been made in developing an effective HIV vaccine. Although an
early HIV vaccine was found to protect against HIV acquisition, the level of protection
provided by this vaccine was relatively low (only a 30% reduction in risk of HIV
acquisition in Thailand), and the vaccine has therefore not been widely used. It is
unlikely that a more effective vaccine will be commercially available before 2030.

While the above prevention programmes are crucial in reducing the risk of HIV at an
individual level, it is also important that interventions be developed with a view to the
social determinants of HIV risk. As discussed previously, factors such as gender
inequality, migration and stigma around HIV/AIDS fuel the growth of the epidemic,
and these need to be addressed.

220
 B. Thembisa HIV and Demographic Model

The information in this section was taken from the website: [Link]

1. Introduction

Many mathematical models have been developed to describe different aspects of the
South African HIV epidemic. Thembisa was developed in an effort to synthesize four
previously-developed models, each of which has different strengths and limitations:

• The Actuarial Society of South Africa (ASSA) AIDS and Demographic model:
This model was developed to evaluate the demographic impact of HIV in South
Africa. Although one of the most detailed demographic models available, it has
not been updated since 2011, and thus does not reflect many of the advances
that have been made in the field of HIV prevention and treatment.

• The STI-HIV Interaction model: This model was developed to assess the role
of different sexual risk behaviours in driving the transmission of HIV and other
sexually transmitted infections (STIs) in South Africa. The model was also
developed to assess the extent to which other STIs have promoted HIV
transmission in South Africa. As with the ASSA model, the model has not been
updated to reflect a number of recent advances in HIV prevention. In addition,
it was not designed to be a demographic model.

• The UCT Paediatric HIV model: This model was developed to assess new
strategies for the prevention of mother-to-child transmission in South Africa, as
well as new strategies for diagnosing and treating paediatric HIV. A limitation
of the model is that it is not integrated into a model of adult HIV transmission,
which means that the effect of adult interventions on mother-to-child
transmission cannot be evaluated dynamically.

• The National Strategic Plan ART Need model: This model was developed to
forecast the number of people who would need antiretroviral treatment (ART),
for South Africa’s 2012-2016 National Strategic Plan (NSP). A key advance
over the ASSA and STI-HIV Interaction models is that it allows for ART
initiation in earlier stages of disease and is based on a CD4 staging system rather
than a clinical staging system. However, the model does not dynamically model
HIV transmission and is not a demographic model.

2. Thembisa Model

Thembisa is a mathematical model of the South African HIV epidemic, designed to

answer policy questions relating to HIV prevention and treatment. Thembisa is also a
demographic projection model and a source of demographic statistics.

“Thembisa” means “give hope” in Xhosa and Zulu. The name reflects the optimism
that has arisen following recent advances in HIV prevention and treatment, and the hope

221
that HIV transmission in South Africa will be substantially curtailed over the next two
decades.

Thembisa incorporates many of the elements of these previously-published models. In

addition it also includes a number of new features not included in any of the previous
models (for example, modelling of male circumcision and pre-exposure prophylaxis).

The Thembisa model has been applied to South Africa as a whole as well as each of the
nine provinces.

A number of versions of the Thembisa model have been published since the publication
of the first version (1.0) in February 2014. The most recent version, Thembisa version
4.6, was published in April 2023 and has been calibrated to recent HIV programme data
(from 2022) as well as results from a national antenatal clinic survey in 2022.

The developers of the Thembisa model are Dr Leigh Johnson (an actuary and
epidemiologist, based at the Centre for Infectious Disease Epidemiology and Research,
at the University of Cape Town) and Prof. Rob Dorrington (an actuary and
demographer, based at the Centre for Actuarial Research, at the University of Cape
Town).

222
 C. HIV prevalence in South Africa

Some key findings from the 2017 National Household Survey conducted by the HSRC
are as follows:

• The survey estimated a total of 7.9 million South Africans living with HIV in
2017.
• The overall prevalence in 2017 (14%) was substantially (and significantly)
higher than that estimated in the previous household survey in 2012 (12.2%).
This translates to about 1.6 million more people living with HIV compared to
the 2012 survey.
• HIV prevalence levels varied substantially by province, from 8.3% in the
Northern Cape, to 18.1% KwaZulu-Natal
• HIV prevalence was higher among women than among men in all age
categories.
• In the age groups 20-24 and 25-29, HIV prevalence among women (15.6% and
27.5%, respectively) was more than double than among men (4.8% and 12.4%,
respectively)
• HIV prevalence in women was highest in the 35-39 age group (39.4%), while
HIV prevalence in men was highest in the 45-49 age group (30.3%).
• An estimated 62.3% of all HIV-positive individuals were receiving
antiretroviral treatment. This is a substantial increase from 31% in [Link]

For further details on the survey, see:

[Link]
=1

223
 PANDEMICS: COVID-19

A. Understanding Covid-19: An introduction for actuaries

This part of the notes was written by Barry Childs and Dave Strugnell (August 2020).

1. Why is it important for actuaries to know about Covid-19?

Actuaries, and the clients that they serve, have a significant interest in anything which
has a material impact on the health and mortality risks of individuals and populations.
Epidemics (which can be defined as outbreaks of an infectious disease which spread
quickly through a given population) are consequently important for actuaries to
understand and respond to, in the form of being able to predict their likely impact. A
pandemic is such an outbreak on a global scale, when the epidemics are not locally or
regionally contained. Their global impact only increases the importance of
understanding their spread and potential impacts. There have been a small number of
disease outbreaks worthy of the designation ‘pandemic’ over the course of the twentieth
and twenty-first centuries, of which the most recent prior to Covid-19 was the (still
ongoing) HIV/AIDS pandemic, about which you will also learn in this course.

Past pandemics in this time frame which are similar in nature to Covid-19, that is,
respiratory illnesses spread by viral transmission, have been severe outbreaks of
influenza viruses: the infamous “Spanish Flu” which killed an estimated 20-50 million
people worldwide between 1918 and 1919, followed by milder influenza outbreaks in
1957-58, 1968 and 2009-10. Covid-19, as you are no doubt all familiar already, is
caused by a coronavirus which is different in molecular structure (in ways that need not
concern us for the purpose of this course) to the influenza virus. There have been past
outbreaks of coronavirus-caused disease which have attracted global attention but had
far less severe impacts than Covid-19: most prominently, the outbreaks now known as
Severe Acute Respiratory Syndrome (SARS) which originated in Guangdong, China in
2002 and killed at least 774 people (648 in mainland China and Hong Kong), and the
Middle Eastern Respiratory Syndrome (MERS) that was first identified in 2012 in
Saudi Arabia, which has resulted in at least 862 deaths worldwide.

Covid-19 dwarfs these past coronavirus outbreaks in mortality and healthcare resource
utilisation impact, though remains mercifully at a considerably lower scale than the
Spanish Flu of a century ago. Furthermore, the responses by citizens to protect
themselves (for example, avoiding public spaces) and the interventions taken by
government to curb the spread of the virus (lockdowns and other measures) have taken
a significant economic toll, with most countries in our globally connected world
anticipating an economic contraction in 2020, with the impact disproportionately felt
by those nations, like South Africa, who were not in the best economic shape going into
the year.

All of the above makes it important for actuaries to have some understanding of Covid-
19. But in addition, the possibility of future pandemics underlines this importance.

224
2. What is SARS-Cov-2 and Covid-19?

HIV is the virus that causes AIDS, the collection of diseases that manifest as a result of
HIV-induced immunosuppression. Similarly, we distinguish between the causal virus
in this instance, Sars-Cov-2, and the disease that results, Covid-19. Although
HIV/AIDS has passed into our everyday language as a neat combination, Sars-Cov-
2/Covid-19 fails to roll off the tongue quite so smoothly, and so typically the term
Covid-19 is employed to cover both the virus and the illness it may lead to. We will
follow this convention in the sections that follow.

The eagle-eyed among you may have questioned why the suffix ‘2’ appears in the virus
name. It is shorthand for Severe Acute Respiratory Syndrome-Coronavirus-2, and you
have probably guessed by now that Sars-Cov-1 is the virus that was responsible for the
SARS outbreak in Asia in the early 2000s. The two coronaviruses share a high degree
of genetic similarity (the details of which are, again, beyond our scope), hence the
naming by the International Committee on the Taxonomy of Viruses (ICTV).

Covid-19 presents in a variety of ways, but respiratory illness and chest infection are
common themes.

3. What do and don’t we know about the spread of the virus and its health impact?

There are two primary pathways of transmission for Covid-19: droplets and aerosol.
The science on the first is very clear, while much uncertainty remains about the second.

Droplets of saliva and respiratory secretions are spread when people cough or sneeze,
as is commonly appreciated, but also in the course of talking and singing. Direct droplet
transmission occurs when those droplets are taken in by the mouth, eyes or nose of a
susceptible (uninfected) individual; indirect transmission takes place when droplets are
brought in by contact of the mouth, eyes or nose with hands which have come into
contact with surfaces on which the droplets lie. The social distancing rule of 1m-1.5m
space between people is an effort to curb direct transmission, and frequent handwashing
or sanitising is prescribed to limit indirect transmission. The encouragement to wear
masks is not so much designed to protect the mask-wearer from infection, but rather to
reduce the infectivity of the mask-wearer by making it less likely that droplets will be
distributed.

Aerosol transmission occurs when droplets remain suspended in the air for some time
and distance, which would obviously negate many of the benefits of social distancing.
The scientific jury is however still out on whether aerosol transmission is likely to be a
significant vector of transmission.

Perhaps the most vexing issue in understanding the spread and impact of Covid-19 is
the wide range of ways in which illness can present in an infected individual. For many
people, perhaps a majority, the experience is entirely asymptomatic, or at least
symptoms so mild that no medication, bed rest or healthcare professional are called for.
At the other extreme, the virus can cause respiratory illness so severe that it requires
specialist hospital intervention (artificial ventilation or the external supply of oxygen)

225
and possibly death. Clarity on the true underlying distribution of illness once infected,
and hence estimation of infection fatality rates (IFRs, i.e. the proportion of people
infected who will die as a result of Covid-19), is plagued by our inability to determine
either the numerator or denominator with complete accuracy:

• deaths that occur due to Covid-19 outside of hospital settings may not be
reported and recorded as such, and thus the official death count statistics will
tend to under-report the true extent of Covid-19 mortality; and
• the impact on the denominator is even more severe: because of testing
constraints and costs, guidelines and protocols in place to determine who ought
to be tested and the fact that a high proportion of those infected will not even
know that they are, given the high incidence of asymptomatic presentation, the
reported cases will undercount true infections by a massive margin (estimates
are that infections may be anywhere between 8x and 25x the number of reported
cases).

We do however have some insights which we can state with confidence. In the first
instance, it appears that children and the young are able to fight off the virus
successfully for the most part, except where there is some underlying comorbidity; at
the other extreme, mortality rates are high amongst the elderly, whose immune systems
are less able to resist. And at any age, the presence of significant comorbidities, in
particular diabetes, hypertension and obesity, as well as HIV (it seems from provisional
studies in South Africa), are all markers for a much higher mortality risk.

4. What treatment and vaccine options are available?

Standard treatment for severely infected patients at the start of the pandemic was
ventilation in an Intensive Care Unit; however, clinical outcomes were not as promising
as had been hoped. The Western Cape was at the forefront of the surge in the pandemic
in our country, and many of the clinical lessons were learned here first. Two significant
strides forward in clinical management, with positive outcomes on health and mortality,
were the provision of High-Flow Nasal Oxygen and the administration of a steroid
called dexamethasone as a first-line treatment. There is however at this stage no
treatment option so reliable that it reduces mortality risk to very low levels.

Work continues on more than 40 vaccine developments around the globe, but at the
time of writing these notes, only one has been approved: Sputnik-V in Russia, about
which numerous safety concerns have been voiced given that it was approved before
entering the final phase of clinical trials. While hopes are high for an effective vaccine,
it seems unlikely that there will be global, affordable availability of a reliable vaccine
before the second half of 2021.

226
 B. ASSA Covid-19 model

1. Brief summary of the global situation, end July 2020

Figure 1 briefly illustrates the journey, at a high level, of the pandemic around the globe.
It initially emerged in Wuhan, China, but has been successfully contained in Wuhan,
the rest of China and, generally, across the rest of Asia. However, it had already spread
to Europe and the USA (especially New York and New Jersey) where it seems to have
spread quietly over the first three months of 2020. The rapidly escalating level of
infections led to a health crisis in the affected regions where health systems were
overwhelmed and a high number of COVID-19 related deaths were recorded. At this
point, strict restrictions on movement and general activity were implemented in most
affected countries in efforts to curb the spread further. Over this period, the number of
deaths fell in both Europe and North America. Both regions are now being watched
closely as they generally ease their restrictions on movement and gatherings.

Other parts of the world had forewarning of the coming pandemic but, even with the
introduction of lockdowns and other non-pharmaceutical interventions (NPI’s), the
pandemic has progressed. South America is the current epicentre of the pandemic with
the majority of the infections and deaths currently occurring across that region. With
the forewarning, the impact has been relatively more gradual but devastating
nonetheless. Africa, on the whole, to date has not been as severely impacted. South
Africa, however, has seen a steady increase in the number of infections and deaths over
May and June, with a rapid escalation over the course of July.

Figure 1: No. of daily COVID-19 confirmed deaths per million lives (by continent) to 21 July
2020 (Shows rolling 7 day average; data source: Our World In Data)

227
2. Model structure

A detailed explanation of the ASSA Covid-19 model would require reference to

mathematical concepts with which you are probably not yet familiar, such as ‘a system
of ordinary differential equations’. We therefore attempt to present the model dynamics
in as intuitive and high-level a way as possible. It should also be pointed out that at the
time of writing, the model is still a work in progress as we add additional functionality
and adapt as the data emerge and our understanding of the pandemic expands.

The underlying model structure is an epidemiologic, compartmental model known as

an SEIR model. Let’s unpack those terms one by one:

1. It’s an epidemiological model in that it is modelling the spread, and

consequences, of an infectious disease;
2. It divides the population into ‘compartments’ with movements between those
compartments over time being modelled; and
3. the basic compartments are those of
a. Susceptible individuals who have not yet been infected;
b. Exposed individuals who are in the incubation period between exposure to
the virus and infection (in which there will be an onset of symptoms, if any);
c. Infected individuals who are able to infect the Susceptible population,
unless they are removed from circulation e.g. through quarantine or self-
isolation; and
d. Removed individuals who have completed the period of infection and are
deemed to be no longer susceptible or infectious.1 The model splits these
into Recovered and Dead populations, who have gotten over or succumbed
to their illnesses respectively.

The basic structure of the model is shown in Figure 2 below, with movement in one
direction, from left to right (of course, there will be some who stay in the S block
throughout the course of the pandemic, never becoming infected; it’s also worth noting
that if there is a risk of reinfection, then there would also be an arrow back from R to
S).

S E I

Figure 2: The basic SEIR compartmental model

1
Early evidence is beginning to emerge of the possibility of reinfection, but this awaits scientific
confirmation; typically with infection diseases, having been infected confers some significant degree of
protection from reinfection.

228
So far, so good. But we know, for example, that the likelihood of hospitalisation and
death varies by age, and so we further split the Infected population as follows:

• in the first instance, by asymptomatic, mild or severe infection;

➢ asymptomatic cases are deemed to be less likely to spread the virus, due
primarily to less coughing and sneezing;
➢ mild cases are those who present with symptoms but will not be
hospitalised; and
➢ severe cases will be admitted to hospital.
• hospitalised cases are further split into general hospital wards and ICU, which
are associated with different mortality levels.

A visual representation of this model structure is shown in Figure 3 below.

Figure 3: ASSA Covid-19 model population structure, July/August 2020

There is of course a great deal of deeper model structure which is beyond the scope of
this course. The most important additional high-level feature of the model to understand
is how we populate the many parameters on which it depends, just some of which are
as follows (don’t get bogged down in the detail; the list is provided merely to give you
a sense of the range of parameters in the model, and is in any event far from exhaustive):

• mean virus incubation period

• mean period of infectiousness
• relative infectiousness of asymptomatic cases relative to symptomatic
• asymptomatic, mild and severe splits, by age band
• lengths of stay in hospital prior to discharge or death

229
 • hospitalisation and mortality rates, by age band
• levels of spreads of the virus during different lockdown conditions, relative to a
baseline with no restrictions or behaviour changes
• the baseline level of spread
• parameters governing heterogeneity in transmission of the virus

In the first instance, we review the scientific literature on an ongoing basis in order to
inform updates to model parameters, and in the second, we analyse emerging evidence
(for example, mortality outcomes in Western Cape hospitals have been key to us
estimating how the mortality curve by age in South Africa may differ from international
experience). This work may inform fixed values for certain parameters which are kept
consistent for all runs, or values which we fix for specific scenarios (for example, for
the most recent release we ran scenarios where the asymptomatic proportion of
infections was set alternately at 35%, 50% and 75%). And then for others, we use this
work to define ranges which the parameter values may take, and then use simulation
techniques (again, beyond scope) to calibrate the values in order to ensure that the
output is consistent with observed data.

At national level, we do not have access to reliable hospital data and hence calibrate the
model results to Covid-19 deaths. We have relied up to the time of writing on the
official reported deaths, but it has become increasingly clear that this is becoming less
and less reliable as the pandemic matures. This is corroborated by the SA Medical
Research Council weekly excess mortality analyses which show excess mortality (i.e.
the difference between actual deaths and those expected in the ordinary course)
significantly higher than official Covid-19 deaths. Their analysis indicates that there
have been more than 39,000 excess natural deaths in South Africa from 6 May to 18
August, which is significantly higher than the official Covid-19 figure of 12,264 for
this period.

it should be noted, however, that not all excess deaths will be directly due to Covid-19;
some will be what are known as collateral deaths occurring as the result of healthcare
resources being diverted from other programmes to combat Covid-19, or individuals
failing to timeously seek treatment for other causes because of the perceived risk of
infection in health facilities. There are however good reasons to suppose that the
majority of excess deaths are directly attributable to Covid-19, and the ASSA modelling
group is on balance of the view that official mortality figures are understated,
particularly in provinces other than the Western Cape. Future model updates will
attempt to make allowance for this under-reporting.

230
3. Other approaches

The compartmental, SEIR model is not the only approach to modelling Covid-19; the
alternatives generally fall under the heading of ‘curve-fitting’, whereby statistical
techniques are used to fit functions to the observed data without concern as to the
mechanics underlying the generation of the observed data (that is to say, without
worrying about how the virus is transmitted or how diseased progresses or is bypassed).
The ASSA working group chose to take the compartmental model route specifically
because it is concerned with those underlying causal dynamics, which allows for
modelling how differently things might turn out if some of the key parameters are
varied, an avenue which is not generally available under curve-fitting approaches.

One such curve commonly used is the Gompertz curve which can be fitted fairly simply
using minimum least squares or maximum likelihood approaches. Curve fitting has the
attraction of being simple to use for well-behaved data, but have the limitation that they
shed little light on disease dynamics or effects of interventions. By way of illustration,
the following figure shows daily reported and 5 day moving average Covid-19 deaths
for the Western Cape, and a fitted Gompertz curve fitted using minimum square error
on 29 August 2020. The approach also requires well behaved data. Data from non-
Western Cape provinces are not as well behaved making such curve fitting approaches
more difficult.

Figure 3: Gompertz curve fitted to Western Cape data

231
4. Model projections

The ASSA Covid-19 model can produce a wide range of outputs for mortality from
different parameter combinations. Given the uncertainty of the disease progression the
modelling group could not settle on a consensus view but nonetheless undertook to
present arrange of reasonable outcomes. The figures below show national daily deaths
and cumulative deaths across a subset of scenarios produced by the model.
700

600

500

400

300

200

100

0
22/04/2020

05/08/2020

18/11/2020
01/04/2020
08/04/2020
15/04/2020

29/04/2020
06/05/2020
13/05/2020
20/05/2020
27/05/2020
03/06/2020
10/06/2020
17/06/2020
24/06/2020
01/07/2020
08/07/2020
15/07/2020
22/07/2020
29/07/2020

12/08/2020
19/08/2020
26/08/2020
02/09/2020
09/09/2020
16/09/2020
23/09/2020
30/09/2020
07/10/2020
14/10/2020
21/10/2020
28/10/2020
04/11/2020
11/11/2020

25/11/2020
02/12/2020
09/12/2020
16/12/2020
23/12/2020
30/12/2020
Figure 4: Daily death projections for South Africa from the ASSA Covid-19 model,
calibrated beginning August 2020

60,000

50,000

40,000

30,000

20,000

10,000

0
01/04/2020

15/04/2020

29/04/2020

13/05/2020

27/05/2020

10/06/2020

24/06/2020

08/07/2020

22/07/2020

05/08/2020

19/08/2020

02/09/2020

16/09/2020

30/09/2020

14/10/2020

28/10/2020

11/11/2020

25/11/2020

09/12/2020

23/12/2020

Figure 5: Cumulative death projections for South Africa from the ASSA Covid-19
model, calibrated beginning August 2020

232
 C. An update
This section provides an update to the above, written in September 2021.

1. Development of the pandemic globally

Covid-19 has persisted globally for over 18 months. Countries have had wide ranging
experience in terms of incidence, hospitalisations and mortality. The pandemic has been
characterized by waves of infections. At the time of writing this update official global
Covid-19 cases and deaths tallied in excess of 227 million and 4,6 million respectively2.
It is widely accepted that official mortality for Covid-19 is understated, and various
attempts have been made to quantify excess mortality or a true death toll of Covid-19.
For example, The Economist estimates the true toll to be 3.3 times the official toll. 3

Most countries continue to manage the disease with non-pharmaceutical interventions

such as social distancing, mask wearing and limitations on gathering size. The broad
consensus on virus spread that has emerged after extensive study is that SARS-COV-2
is mostly spread via aerosol making enclosed crowded spaces in particular risky.

Since the start of the pandemic scientists have been working on possible vaccines. A
number of vaccines were approved for use by medicine authorities around the world
and an extensive rollout of vaccine programs from late 2020. Vaccine distribution
became a topic of debate with regard to equity as wealthy countries secured large
amounts of vaccines for themselves early, leaving little for poorer countries. Vaccines
have been shown to be safe (with minimal side effects) and effective for the prevention
of severe disease and death. One key debate of vaccine programs in times of limited
supply of stock has bene how to prioritise – who to give the vaccines to first.
Logistically vaccine programs have had challenges since most Covid-19 vaccines
require two doses. Vaccine hesitancy has been a particular issue for Covid-19 vaccines
with social media amplifying anti-vaccine sentiments and fears.

Another theme that emerged during the ongoing pandemic were so called variants of
concern. There were mutated SARS-COV-2 viruses that gained different characteristics
with respect to infectivity and severity. Viruses mutate all the time as part of their
survival mechanism, but certain strains came to dominate the spread of Covid-19. The
so called Delta variant, discovered by South African scientists became a dominant strain
globally and was shown to be more infectious than previous strains. A key concern for
variants was the prospect of immune escape – the ability for a variant to reinfect a
previously infected individual or a vaccinated individual nullifying the immune
response. So far this risk has not borne out in practice to any notable degree.

2
[Link]
3
[Link]
[Link]

233
2. Development of the pandemic in South Africa

South Africa’s covid-19 experience has, so far, been far worse than initial modelled
projections and estimate. At the time of writing, confirmed mortality is near 90 000 and
excess mortality of near 260 000 natural deaths (of which the SAMRC estimates 85-
95% are due to covid-19 due to the pattern of excess deaths being closely aligned to
covid-19 case numbers). We have gone through three waves of infections coinciding
with winter 2020, summer 2020/21 and winter 2021, with the third wave lingering
longer than the previous two. The third wave has been the most severe. Concerns linger
over a possible fourth wave although the general view is that it be less severe in terms
of hospitalization and mortality because of the proportion of the population already
exposed and the vaccine campaign which has been rolling out since May 2021.

Source: [Link]

234
Source: [Link]

Excess mortality has also emerged in reported life insurance statistics. ASISA released
a report4 citing 1 023 083 death claims between 1 April 2020 and 31 March 2021, an
increase of 309 733 death claims compared to the statistics for the previous 12 months
when 713 350 claims were received. Note that these do not represent individual counts
but policy counts (with some individuals having more than one policy), but nonetheless
the increase over the previous year affirms much higher mortality over the period.

The vaccine programme in South Africa got off to a slow start due to difficulties in
securing budget for procurement then difficulties in securing stock from manufacturers
due to global demand exceeding supply. South Africa’s medicine regulatory authority
SAHPRA approved the use of the Pfizer (2 dose) and Johnson & Johnson (1 dose)
vaccines for use in South Africa. Healthcare workers were prioritised to get vaccines
first given their risk of exposure and the importance of keeping the healthcare
workforce productive during covid-19 waves in order to treat patients. For the
remainder of the population triage was based on age, with over 60s being asked to get
vaccinated first, with the program opening up successively to younger age groups over
time. At the time of writing all adults are eligible for a vaccine and the program is
delivering close to 1 million doses of vaccine every four days between the private and
public sectors, with the private sector delivering about a third of vaccine doses.

Vaccine coverage has been higher in urban areas and among the medically insured (on
a medical scheme). As at 5 September 2021 50% of over 50’s had had at least one
vaccine (some J&J some Pfizer) and 39% were fully vaccinated.

4
[Link]
more-than-a-million-death-claims-in-12-month-period/

235
Figure: % of the over 50 population with at least 1 vaccine dose by sub district
Source: DOH daily vaccine statistics

Of concern is the observable hesitancy across all age groups but especially younger age
groups who one might surmise consider themselves at lower risk of severe covid-19.
The Department of health's targets are to vaccinate at least 70% of the adult population
by the end of the year. Vaccine take-up among over 60s for first doses appears to be
plateauing closer to 60%. The drop off in demand for vaccines has been faster in
younger age groups leading to concerns those groups will plateau at even lower levels.

Figure: % of age group with at least 1 vaccine dose by vaccine program week
Source: DOH daily vaccine statistics

236
3. Modelling

ASSA’s covid-19 working group did not release a further update of the covid-19
projection model mainly due to capacity constraints of the volunteers. The focus of the
group shifted to monitoring and assisting stakeholders deal with various aspects of the
vaccine programme. The best local models are those maintained by Louis Rossouw
found at:
[Link]
19_in_south_africa_at_a_provincial_level.html
and the official South African Covid-19 Modelling Consortium found at:
[Link]

The ASSA model was not designed to deal with the waves observed in the pandemic,
in part because it is still not known precisely what drives these waves. The effect cannot
be seasonal as waves came in both summer and winter. Nor can they be solely
attributable to lockdown or other NPIs as there are enough counter examples to question
the degree of the effect. That is not to say that social distancing, mask wearing and other
NPIs are not effective, just that it is difficult, based on the data we can observe, to
attribute the wave phenomenon to those effects alone.

ASSA’s second model was designed as a tool to assess the likely effect of a wide range
of variables such as the R0 of Covid-19, the proportion of cases that would be
asymptomatic, etc. the nature of the compartmental model is such that the virus spreads
through the susceptible population based on the input parameters. Model outputs
typically showed very high levels of infected persons with varying hospitalization and
mortality projections depending on the range of inputs selected.

Overall, what has come to pass has been more severe than most of the those involved
in the group anticipated, and while the ASSA model did not allow for waves, we have
nonetheless reached high proportions of the population having been infected, most with
mild or asymptomatic cases.

South Africa’s age profile would suggest an infection fatality rate of 0.3%. However,
given South Africa's burden of disease, we might permit a higher figure of say 0.5% to
be used. Using this figure, and 85% of excess deaths as the true mortality figure, an
estimated 43,8 million South Africans have been infected to date (74% of the
population).

4. Concluding remarks

While much uncertainty still remains we trust that the outline above gives you some
sense of the issues and the approach that has been taken to modelling Covid-19, and
leave you with final words from the statistician George Box, always worth bearing in
mind but especially when attempting to model something as complex and uncertain as
this: “All models are wrong, but some are useful.”

237