Boston artist Steve Mills - realistic painting

Friday, April 9, 2010

South Africa AIDS statistics

The statistics discussed here come from two prevalence studies that estimate how many people are living with HIV in South Africa, and two reports on AIDS deaths. Viewed together these sources give an idea of the scale of South Africa's HIV epidemic. If you are looking for statistics from elsewhere, try ourstatistics section.
The first section is based on the report of the Department of Health "National HIV and Syphilis Sero-prevalence Survey in South Africa 2007", published in 2008. This annual study looks at data from antenatal clinics and uses it to estimate HIV prevalence amongst pregnant women.
The second section is based on the report of the "South African National HIV Prevalence, HIV Incidence, Behaviour and Communication Survey, 2008". In this survey, a sample of people were chosen to represent the general population. Of those who were eligible, 64% agreed to give a blood sample to be anonymously tested for HIV. The report contains estimates of HIV prevalence in various groups of people, derived from this general population sample.
Seen together, the two prevalence studies provide a clearer picture of the South African epidemic than either of them viewed alone.
The third section looks at AIDS-related deaths using data from death certificates. Reports published by Statistics South Africa contain the raw data, while the article "Identifying deaths from AIDS in South Africa" analyses a large sample of death certificates and attempts to estimate how many deaths caused by HIV have been misclassified.
The page goes on to compare the two types of prevalence study and to draw conclusions.

The South African Department of Health Study, 2007

Based on its sample of 33,488 women attending 1,415 antenatal clinics across all nine provinces, the South African Department of Health Study estimates that 28% of pregnant women were living with HIV in 2007. The provinces that recorded the highest HIV rates were KwaZulu-Natal, Mpumalanga and Free State. The Northern Cape and Western Cape recorded the lowest prevalence.
Until 1998 South Africa had one of the fastest expanding epidemics in the world, but HIV prevalence now appears to have stabilized, and may even be declining slightly. Among teenage girls, the rate fell from 16.1% in 2004 to 12.9% in 2007, possibly indicating a drop in the rate of new infections. The health department believes this is due to a change in safer sexual practices among younger women. The inability to moderate cultural circumstances is believed to be a factor in the high and rising HIV prevalence among relatively older women.
More historical prevalence figures can be found in our AIDS in South Africa page.

Estimated HIV prevalence among antenatal clinic attendees, by province

Province2001 prevalence %2002 prevalence %2003 prevalence %2004 prevalence %2005 prevalence %2006 prevalence %2007 prevalence %
Free State30.128.830.129.530.331.133.5
North West25.226.229.926.731.829.029.0
Eastern Cape21.723.627.128.029.528.626.0
Northern Cape15.915.116.717.618.515.616.1
Western Cape8.612.413.115.415.715.112.6

Estimated HIV prevalence among antenatal clinic attendees, by age

Age group (years)2001 prevalence %2002 prevalence %2003 prevalence %2004 prevalence %2005 prevalence %2006 prevalence %2007 prevalence %
Because infection rates vary between different groups of people, the findings from antenatal clinics cannot be applied directly to men, newborn babies and children. This is why South Africa has sought also to survey the general population.

The South African National HIV Survey, 2008

The National HIV Survey is a "household" survey. This involves sampling a proportional cross-section of society, including a large number of people from each geographical, racial and other social group. The researchers take great pains to try to make the sample as representative as possible, and the findings are later adjusted to correct for likely over- or under-representation of individual groups (according to census data).
The survey's fieldworkers visited 15,000 households across South Africa, of which 13,440 (90%) took part in the survey. Of the 23,369 people within these households who were eligible to take part, 20,826 (89%) completed an interview and 15,851 (64%) agreed to take an HIV test.
Based on this survey, the researchers estimate that 10.9% of all South Africans over 2 years old were living with HIV in 2008. In 2002 and 2005, this figure was 11.4% and 10.8%, respectively, showing a degree of stabilisation. Among those between 15 and 49 years old, the estimated HIV prevalence was 16.9% in 2008. The survey found the prevalence among children aged 2-14 to be 2.5%, down significantly since 2002, when prevalence was 5.6%.

Estimated HIV prevalence (%) among South Africans aged 2 years and older, by age, 2002-2008

Children (2-14 years)
Youth (15-24 years)9.310.38.7
Adults (25 and older)15.515.616.8
15-49 year olds15.616.9216.9
Total (2 and older)11.410.810.9

Estimated HIV prevalence among South Africans, by age and sex, 2008

Age Male prevalence %Female prevalence %
Among females, HIV prevalence is highest in those between 25 and 29 years old; among males, the peak is in the group aged 30-34 years.

HIV prevalence (%) by province 2002-2008

Free State14.912.612.6
North West10.310.911.3
Eastern Cape6.68.99.0
Northern Cape8.45.45.9
Western Cape10.71.93.8
The results of this study suggest that KwaZulu-Natal, Mpumulanga and Free State have the highest HIV prevalence. However, the relatively small sample sizes may limit precision, and in several cases the ranges of uncertainty overlap.

HIV prevalence by population group, 2008

Population groupPrevalence (%)

Studies of AIDS deaths

All reported deaths

In October 2008, Statistics South Africa published the report "Mortality and causes of death in South Africa, 2006". This large document contains tables of how many people died from each cause according to death notification forms.
The report reveals that the annual number of registered deaths rose by a massive 91% between 1997 and 2006. Among those aged 25-49 years, the rise was 170% in the same nine-year period. Part of the overall increase is due to population growth. However, this does not explain the disproportionate rise in deaths among people aged 25 to 49 years. In 1997, this age group accounted for 29% of all deaths, but in 2006 it accounted for 42%.

Reported deaths from all causes, 1997 to 2006

Year of deathAge (years)Total
Increase 1997-200693%72%170%54%-79%91%
The influence of population growth can be removed by looking at death rates per 100,000 people, which are provided by Statistics South Africa in another report called "Adult mortality (age 15-64) based on death notification data in South Africa: 1997-2004". These data show that between 1997 and 2004, the death rate among men aged 30-39 more than doubled, while that among women aged 25-34 more than quadrupled. The changes are even more pronounced when deaths from natural causes only are examined. Over the same period there was relatively little change in the death rates among people aged over 55 and those aged 15-20. In their report, Statistics South Africa call such developments "astounding", "alarming" and "disturbing".


In 2006, HIV was recorded as a cause of death in only 14,783 cases. However, according to researchers from the Medical Research Council of South Africa (MRC), this figure is a massive underestimate, because the majority of deaths due to HIV are misclassified.
People whose deaths are caused by HIV are not killed by the virus alone, but HIV should be recorded as an underlying cause if it "initiated the chain of morbid events leading directly to death". In other words, if someone contracts tuberculosis and dies from it because their immune system has been weakened by HIV then HIV should be included among the underlying causes. The MRC researchers claim that in many cases, this does not happen; instead, the doctor records only the immediate cause of death such as tuberculosis or respiratory infection. This could be because the doctor does not know the deceased person's HIV status. Alternatively, they may seek to conceal HIV infection to spare stigmatisation of relatives, or to avoid invalidating life insurance claims. As The Lancet notes, authorities are largely to blame:
“Social stigma associated with HIV/AIDS, tacitly perpetuated by the Government's reluctance to bring the crisis into the open and face it head on, prevents many from speaking out about the causes of illness and deaths of loved ones and leads doctors to record uncontroversial diagnoses on death certificates.... The South African Government needs to stop being defensive and show backbone and courage to acknowledge and seriously tackle the HIV/AIDS crisis of its people.”1
The MRC team analysed a 12% sample of death certificate data from the year 2000-2001, and compared it to all the data from 1996. When they looked at deaths for which HIV was a reported cause, they saw that rates (deaths per thousand) had increased according to a distinctive age-specific pattern. The greatest increases were in the age groups 0-4 and 25-49 years, while death rates among teenagers and older people remained more or less unchanged.
The researchers observed that nine other causes of death had increased substantially according to the same distinct age pattern as HIV. They then estimated how much of the increases were likely to be caused by HIV, and concluded that 61% of deaths related to HIV had been wrongly attributed to other causes in 2000-2001. In adults, tuberculosis accounted for 43% of misclassified deaths, and lower respiratory infections for another 32%. Among infants, most of the excess deaths had been misclassified as lower respiratory diseases or diarrhoeal diseases. According to the MRC results, HIV caused the deaths of 53,185 men aged 15-59 years, 59,445 women aged 15-59 years, and 40,727 children under 5 years old in the year 2000-2001.
The MRC estimates come very close to those made by a computer model of the Actuarial Society of South Africa, called ASSA2003. According to ASSA2003 calculations, HIV caused 108,170 deaths in 2000 and 147,525 deaths in 2001.
Statistics South Africa have analysed the MRC study and found that its methods and conclusions are generally sound.

Other recent estimates

The head of the MRC has stated that AIDS killed around 336,000 South Africans between mid-2005 and mid-2006.2
The ASSA2003 provincial model calculates that 345,640 people died because of AIDS in 2006 - comprising 47% of all deaths. Among adults aged 15-49 years, it estimates that 71% of all deaths were due to AIDS.3
UNAIDS/WHO estimate that AIDS claimed 350,000 lives in 2007 - nearly 1,000 every day.4

Comparing the prevalence studies

It is possible to compare the results of the National HIV Survey 2008 with those of the Department of Health Study 2005.
HIV prevalence according to the Department of Health Study 2005:
  • 29.1-31.2% amongst antenatal clinic attendees (30.2% is the best estimate)
HIV prevalence according to the National HIV Survey 2008:
  • 10.0-11.9% in the whole population (10.9% is the best estimate)
  • 15.5-18.4% amongst all people aged 15-49 years old (16.9% is the best estimate).
The rates found among pregnant women are significantly higher that those found among all adults - so why could this be?

Limitations of the Department of Health Study

Antenatal surveillance is internationally recognised as the most useful way of assessing HIV prevalence in countries with generalised epidemics. Pregnant women are sexually active and constitute an easily identifiable, accessible and stable population. They are more likely than any other single group to be representative of the general adult population. Nevertheless, there are a number of limitations to the Department of Health's technique.
The greatest difference between the two studies concerns prevalence among women aged 15-19 years old, for which the antenatal survey produces a rate much higher than the household survey (15.9% compared to 9.4%). This is, at least in part, probably because not all young women are sexually active, and those represented in the antenatal data are by definition engaging in unprotected sex, which puts them at higher risk of HIV infection. Overestimation of HIV prevalence in this age group is a known bias in antenatal studies.
It is possible that overestimation occurs in older age groups as well, particularly as those who use condoms or abstain from sex stand less chance of both HIV infection and pregnancy. On the other hand, underestimation might also occur: for example, studies have shown that HIV lowers fertility.

Limitations of the National HIV Survey

The advantage of the National HIV Survey is that it can give a better idea of HIV prevalence levels among men, children and non-sexually active women. The survey also recorded a vast amount of other data besides the age and location of respondents (most of which is beyond the scope of this page), including information on race, wealth and education. Participants were also interviewed about factors that might influence their risk of HIV infection, such as behaviour, knowledge and risk awareness.
Although the study attempted to survey as representative a population sample as possible, it recognises that some groups were excluded. Only people living in homes or hostels were contacted, so there was no representation of homeless people and those living in police and army barracks, prisons, hospitals and educational institutions. This probably resulted in underestimation of some prevalence figures. Additionally, by excluding all children below 2 years of age (because they cannot be reliably tested for HIV using antibody tests), the survey missed a significant proportion of children who acquired HIV from their mothers. The survey's design also meant some groups that may be of particular interest for the understanding of the epidemic could not be captured in sufficient numbers, including men who have sex with men, injecting drug users and sex workers.
The survey had also had a fairly high rate of non-response with just 64% of all eligible participants agreeing to an HIV test. The effect of non-response on accuracy is uncertain. It is difficult to conclude whether those who refuse to be tested are more or less likely to have HIV. The only certain effect of the low response rate is that it increases uncertainty.5
The National HIV Survey is the the third of its kind to be conducted across the whole of South Africa.

Conclusion of the comparison

Neither prevalence study sets out to mislead or to contradict the other. Each uses a standard surveillance technique and clearly explains all of its methods and calculations. Most of the observed differences are the result of choosing different groups of people to be tested, since these groups differ in how well they are able to represent the general population.
In such a large and diverse country as South Africa, no-one can know exactly what the true figures are. What is essential is that the limitations of each study are acknowledged whenever their results are interpreted. To illustrate why this is so important, this page has suggested a few reasons why the figures might vary, though this is by no means an exhaustive list.
UNAIDS and WHO recommend that antenatal and population-based studies should both be conducted at regular intervals. In countries with generalised epidemics, antenatal clinic attendees are thought to represent the adult population with good accuracy. Moreover, when conducted regularly such surveys can reveal long-term trends in prevalence. On the other hand, household surveys tell us more about the nature of the epidemic by providing prevalence data according to gender, race, wealth and other characteristics. Such information informs better interpretation of antenatal data.

National estimates based on all surveys

Based on a wide range of data, including the household and antenatal studies, UNAIDS/WHO in July 2008 published an estimate of 18.1% prevalence in those aged 15-49 years old at the end of 2007. Their high and low estimates are 15.4% and 20.9% respectively. According to their own estimate of total population (which is another contentious issue), this implies that around 5.7 million South Africans were living with HIV at the end of 2007, including 280,000 children under 15 years old.6
The ASSA2003 model produces a similar estimate of 5.4 million people living with HIV in mid-2006, or around 11% of the total population. It predicts that the number will exceed 6 million by 2015, by which time around 5.4 million South Africans will have died of AIDS.7


What is clear from every study is that there is an exceptionally severe epidemic of HIV/AIDS in South Africa. This epidemic affects all parts of the population, though women are more likely to be infected than men. Many tens of thousands of people are dying.

No comments:

Post a Comment