Review of the HSRC’s prevalence, incidence, behaviour and communication survey
2 April 2009. Related: Other news.
Nathan Geffen, TAC
The Human Sciences Research Council (HSRC) published its much awaited 2008 HIV prevalence, incidence, behaviour and communication Survey in June 2009.  This is the third such survey. The previous ones were conducted in 2002 and 2005. Previously the survey only included people over two years old, but this one included all ages.
This year’s report is much shorter than the other two and contains less detail. The HSRC explains that more detail will be made available in journal articles that will be submitted for publication. The report contains a plethora of interesting data.
The sample consisted of 15,000 households, 15 from each of 1,000 enumeration areas. Of these 13,440 (90%) were valid occupied households. The occupants of 10,856 (81%) agreed to be interviewed.
23,369 individuals were identified as eligible to be interviewed (a parent or guardian was interviewed for children aged 11 and under). 20,826 (89%) completed a behavioural interview. 15,031 (64%) agreed to provide a blood sample to be tested for HIV. The HIV tests were anonymously linked to the interview. The acceptance rates for HIV tests by sex were 62% for males and 69% for females, and by race were 69% for Africans, 53% for whites, 75% for coloureds and 48% for Indians.
The researchers extensively analysed factors associated with refusal to have an HIV test that could have biased the results. They concluded, “Although some associations were statistically significant due to the large sample sizes, the differences between those tested and not tested were all less than 10% and most were less than 5%. Based on this more detailed analysis of HIV risk-associated characteristics in survey respondents who were interviewed and tested and those who were interviewed but refused HIV testing we conclude that the HIV survey results were not biased due to HIV testing refusal.”
At most four people from any household were interviewed, one from each of these age groups: under 2, 2-14, 15-24, 25 and above. Calculations were weighted to compensate for under-representation of enumeration areas, households and individuals.
The survey does not include people in institutions, e.g. prisons, army barracks, university residences and boarding schools.
The weighted HIV prevalence rate was 10.6%. No confidence interval was provided for this estimate. For people over the age of two there was no significant difference in prevalence at a national level from 2002 to 2008:
- 2002: 11.4% (95%CI 10.0-12.7)
- 2005: 10.8% (95%CI 9.9-11.8)
- 2008: 10.9% (95%CI 10.0-11.9)
The consistency across all three surveys implies that we can say with confidence that prevalence among people in South Africa over the age of two who are not living in institutions is approximately 11% and that it has not changed significantly in recent years.
At a provincial level, there is only one significant change in prevalence. In 2002, the Western Cape prevalence was 10.7% (95%CI 6.4-15.0). This changed to 1.9% (95%CI 1.2-3.0) and 3.8% (95%CI 2.7-5.3) in 2005 and 2008 respectively. As can be seen, the change from 2002 to 2005 is significant but not from 2005 to 2008. There is no reason why the prevalence in the Western Cape should have dropped so much from 2002 to 2005, even taking into account the province’s successful prevention of mother-to-child transmission programme (PMTCT) programme, whose effect on prevalence in any case would be partially or wholly offset by the corresponding success of the HAART programme.  Furthermore the Western Cape antenatal HIV prevalence increased from 12.4% to 15.7% from 2002 to 2005. Most likely, this is a sampling problem or statistical anomaly in one or more of the three HSRC surveys.
When broken down by the three older age groups, there is a significant drop in prevalence in the 2-14 age group from 2002 to 2008. This possibly reflects some effectiveness of the PMTCT programme. There was no significant change in the 15-24 and 25+ age groups across the three surveys. See Table1.
Table 1: HIV prevalence by province for people aged above two years old in the three HSRC household surveys. Percentages are weighted. Taken from page 32 of the 2008 report.
The 2008 survey also introduced prevalence measures for people in groups at high risk of HIV infection, an important development, which will hopefully provide a rich source of data. See Table 2.
Table 2: HIV prevalence in groups at risk
It is worth noting that 4,238 males over 15 agreed to have an HIV test. Of these 86 affirmed that they have sex with other men, ie 2%. Taken from page 36 of the 2008 report.
The HIV/AIDS National Strategic Plan was published in April 2007. It set a target to reduce HIV incidence by 50% by 2011. Consequently the HSRC report has a detailed discussion on incidence.
The 2005 survey used the BED-assay to estimate incidence. In December 2005 a UNAIDS reference group released a statement on the reliability of this method. The group explained, “Based on the … evidence, the Reference Group recommends that at present the BED-assay not be used for routine surveillance applications, neither for absolute incidence estimates, nor for monitoring trends.”  The BED-assay methodology for estimating incidence has subsequently been improved, though it is complex and should ideally be corroborated with additional studies. 
The BED-assay calculation of incidence was not ready at the time the 2008 survey went to print. Instead the survey determined incidence for each year of age for 15 to 20 year-olds by deriving it from single year age prevalences. It concluded that, “there was a substantial decrease in incidence in 2008 in comparison to 2002 and 2005, especially for the single age groups 15, 16, 17, 18, and 19.” This is an unusual incidence calculation method that works as follows (using an example in the report):
- Proportion of 14-year olds infected is 0.0311 (ie prevalence is 3.11%).
- Proportion of 15-year olds infected is 0.0389 (ie prevalence is 3.89%).
- The difference in prevalence between 14 and 15-year-olds is 0.0389 – 0.0311 = 0.0078.
- The proportion of the population of 14-year-olds that is uninfected is 1 – 0.0311 = 0.9689.
- Incidence is calculated as the change in prevalence divided by the proportion of the population at risk, ie 0.0078/0.9689 = 0.008 (ie incidence for 15-year olds is 0.8%).
The calculation assumes that incidence remains the same from one calendar year to the next and that prevalence in the 15-20 year age group is unaffected by AIDS (hence the calculation cannot be done for older ages). The validity of these assumptions is uncertain. More critically, the method does not calculate confidence intervals or p values. The report also does not provide confidence intervals for prevalence for each year of age 15 to 20. However, given that the national 95% confidence intervals for prevalence in 15-24 year olds are wide and overlapping across the three surveys (7.5%-11.4% in 2002, 8.7%-12.0% in 2005 and 7.2%-10.4% in 2008), it is unlikely that any incidence calculation for 15-20 year-olds based solely on prevalence estimates would show a statistically significant decline.
It is plausible that incidence has declined from 2002 to 2005 to 2008. However, there is insufficient evidence to conclude this from the HSRC report.
Unless more compelling data is published by the HSRC it will be difficult to assess whether the incidence target of the NSP has been achieved. Hopefully the BED-assay results will provide a robust estimate of incidence. If not then researchers should consider doing large longitudinal surveys in some of the country’s high-risk areas, so that we can better understand incidence. One recent such study found an incidence of 3.4 per 100py (95%CI 3.1-3.7) in Umkhanyakude district, Kwazulu-Natal and no sign of decline over a five year period. 
The report examined these behavioural determinants of HIV incidence: sexual debut, intergenerational sex, multiple sexual partners and condom use. In contrast to the prevalence measure, this part of the survey depends on the manner in which questions are phrased, how well they are understood and the willingness of interviewees to tell the truth.
Nationally, 5% (95%CI 3.8-6.5) in 2002, 8.4% (95%CI 7.2-9.9) in 2005 and 8.5% (7.1-10.1) in 2008 of 15 to 24 year-olds said they had sex before reaching 15 years old. While there is a significant difference between 2002 and the 2005 to 2008 period, it is difficult to interpret this. At provincial level, the increase between 2002 and the 2005-2008 period was only significant in North West and Free State.
In 2005 2% of males (95%CI 1.0-4.2) and 18.5% (95%CI 13.7-24.4) of females in the 15-19 age group said they had a sexual partner at least five years older than them. In 2008 0.7% (95%CI 0.2-2.7) of males and 27.6% (95%CI 21.7-34.5) of females said they had a sexual partner at least five years older than them. This was not measured in 2002. The differences between 2005 and 2008 are not statistically significant.
Among males aged 15 to 49, 9.4% (95%CI 8.1-10.9) in 2002, 17.9% (95%CI 15.5-20.6) in 2005 and 19.3% (95%CI 17.3-21.6) in 2008 said they had more than one partner in the past 12 months. Amongst females aged 14 to 49, 1.6% (95%CI 1.1-2.3), 2.9% (95%CI 2.3-3.7) and 3.7% (95%CI 2.9-4.8) made the same claim. It is not clear why there was a large and significant rise from 2002 to 2005. (Tables 3.15 and 3.16 of the HSRC report appear to give contradictory data for 15 to 49 years olds nationally in 2002. I have assumed Table 3.16 has an error.)
It is plausible that both these factors, ie high rates of intergenerational sex of female youth and high rates of multiple partners of males are drivers of the HIV epidemic.
The survey reports significantly increased reported condom use overall, by sex and by age group between the 2005 and 2008 reports. Reported condom use at last sex was 35.4% (95%CI 33.4-37.3) in 2005 and 62.4% (95%CI 60.2-64.6) for people 15 years and older. Reported condom use was similar between males and females. It is unclear how condom use could have increased so massively over a three year period. Perhaps the increase reflects reality, but it could also be that it has become more socially desirable to report use of condoms.
Amongst people who reported multiple sexual partners, reported condom usage was 75.2% (95%CI 69.2-80.4), with no significant change since 2002 and 2005. There was also no significant difference between males and females.
Awareness of HIV status
The survey found that the percentage of people who reported having had an HIV test in the last 12 months and knowing the results doubled between from 11.9% in 2005 to 24.7% in 2008. More women said they know their status than men (28.7% v. 19.9%). These results are significant, though worryingly low and probably explain why so many people present so late for HAART. Among groups defined as being at high risk, 35.7% (95%CI 32.5-39) of African women aged 20-34 years said they had been tested in the last 12 months and learnt their status. Next were MSM, but the confidence interval was very wide (95%CI 17.2-40.3). For African men aged 25-49 it was 25% (95%CI 21.6-28.7).
Knowledge of HIV
The survey measured knowledge of HIV prevention. If interviewees agreed with these two statements they were scored as knowledgeable about HIV prevention:
- To prevent HIV infection, a condom must be used for every round of sex
- One can reduce the risk of HIV by having fewer sexual partners
The survey reports large and significant declines in this measure since 2005 in people aged 15 to 49 years. Participants who answered yes to both questions declined from 64.4% (95%CI 62.5-66.3) in 2005 to 44.8% (95% CI 42.9-46.7) in 2008.
This is implausible. Knowledge is unlikely to deteriorate in what is essentially the same population over a three year period and certainly not by such a large amount. On the contrary it is only likely to increase. Much more plausible explanations are: this is a statistical anomaly; there is something different about the samples between the 2005 and 2008 surveys; the HIV knowledge questions were asked differently; or there was a something different about the way the interviewers carried out their functions.
Furthermore, the wording of the two test statements could have been better; they are not a convincing measure of HIV prevention knowledge. For example, with regard to the first statement, interviewees might consider having an undetectable viral load sufficient to prevent HIV infection. With regard to the second statement interviewees might believe that using condoms consistently negates the risk of having multiple partners. Even if these views are incorrect (and it is not clear that they are), they imply a sophisticated knowledge of HIV prevention.
Participants had to correctly indicate whether these statements were true or false to be marked as knowledgeable about AIDS myths:
- There is a cure for AIDS
- AIDS is caused by witchcraft
- HIV causes AIDS
- AIDS is cured by having sex with a virgin
63.8% (95%CI 62.5–65.1) of people got all four correct. This was not significantly different from 2005.
(Note: The survey also included measures of the exposure of four HIV communication programmes, but I have omitted these in this report.)
1. Shisana et al. South African national HIV prevalence, incidence, behaviour and communication survey 2008: A turning tide among teenagers? 2009. Cape Town: HSRC Press.
2. Azevedo V. City Health: Khayelitsha. July 2007.
3. UNAIDS. UNAIDS Reference Group on Estimates, Modelling and Projections statement on the use of the BED-assay for the estimation of HIV-1 incidence for surveillance or epidemic monitoring. December 2005.
4. CDC. Using the BED HIV-1 Capture EIA Assay to Estimate Incidence Using STARHS in the Context of Surveillance in the U.S.
5. Bärnighausen T. Lack of a decline in HIV incidence in a rural community with high HIV prevalence in South Africa, 2003-2007. AIDS Res Hum Retroviruses. 2009 Apr. 25(4) 405-9.