The 19 countries that have conducted national population based surveys are Burkina Faso, Burundi, Cameroon, Congo, Equatorial Guinea, Ghana, Guinea, Kenya, Lesotho, Mali, Niger, Republic of South Africa, Rwanda, Senegal, Sierra Leone, Tanzania, Uganda, Zambia, and Zimbabwe (table 1).12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31
Most surveys have targeted the adult population (typically women aged 15–49 years and men aged 15–49 or 15–59 years) except in Zimbabwe31
and the RHRU study in South Africa24
where the focus was on young people (aged 15–29 and 15–24 years respectively), and the Burundi survey (including all people older than 12 years).13
Two surveys, the South African Human Sciences Research Council (HSRC) household survey and the Uganda AIS survey also included children.25,29
While some surveys have added HIV testing to a pre‐existing standard methodology (notably the international DHS survey programme), other surveys were specifically conducted to collect information on HIV and AIDS—for example, the AIS and many of the surveys that are not part of an international survey programme. The survey in Congo15
was limited to urban areas only.
Table 1Characteristics of population based surveys
There was large variation in the sample sizes of surveys, ranging from less than 1500 people in Equatorial Guinea to more than 15
000 in the HRSC survey in South Africa and the Uganda AIS (table 1). While one would expect larger sample sizes in countries with lower prevalence, the sample sizes were not related to the expected HIV prevalence—for example, in those countries with similar population sizes, the sample size in the high prevalence countries Zimbabwe (10
744) and Zambia (3807) was similar to the sample size in the low prevalence countries Senegal (7524) and Mali (6846). In most of the surveys the biological specimen collected for HIV testing were dried blood spots (DBS) from capillary blood, while in four surveys venous blood was drawn (in the Zambia survey DBS were prepared from the venous blood). Only the survey among young people in South Africa used oral fluids.24
Although the first few DHS surveys in Mali in 2001 and Zambia in 2002 did not link HIV results to the sociodemographic and behavioural information, all other surveys were linked (table 1).
Household response rates were high in most countries: Burkina Faso (99.4%), Cameroon (97.3%), Congo (>95%), Ghana (98.7%), Guinea (99.2%), Kenya (96.3%), Lesotho (94.9%), Mali (97.8%), Rwanda (99.7%), Senegal (98.5%), Tanzania (98.5%), and Uganda (96.8), Zambia (98.2%), and Zimbabwe (95%). However, three surveys had household response below 90%, including those for South Africa (84.1% and 88.3% for its two surveys respectively) and Equatorial Guinea (75.4%). Not enough information was provided to derive the household response rates for the surveys in Burundi, Niger, and Sierra Leone.
At the individual level, countries that reported relevant information show a clear pattern of higher response rate among women compared to men, and in urban compared to rural areas (table 2), The overall HIV testing rate of the populations surveyed varied between the lowest values of 68.2% for women and 62.2% for men in the South African surveys to almost 97.3% for women and 95.4% for men in Rwanda. HIV testing rates were below 70% for women in South Africa, and for men in Equatorial Guinea, Lesotho, and South Africa. Not enough information was provided on testing rates for Burundi, Congo, Niger, and Sierra Leone. No specific information on absenteeism and refusal rates was available for Burundi, Congo, Mali, Niger, South Africa young people survey, Sierra Leone, and Zimbabwe. Absenteeism among women varied from 0.2% in Guinea to 6.0% in Kenya and 19.1% in Equatorial Guinea. Refusal of the HIV test among women varied from 0.3% in Equatorial Guinea to 14.4% in Kenya, 15.7% in Zambia, and 30.2% in the South African HSRC survey. Absenteeism among men was higher than among women and varied from 0.4% in Guinea to 12.2% in Kenya and 29.5% in Equatorial Guinea. Refusal of the HIV test among men varied from 1.1% in Equatorial Guinea to 16.6% in Lesotho 34.6% in the South African HSRC survey.
Table 2National population surveys: individual response rates and HIV prevalence
The results of the available national population based surveys show the extreme variation of HIV prevalence in sub‐Saharan Africa (table 2). HIV prevalence among adults varied from below 1% in Niger and Senegal to 23.5% in Lesotho, reflecting a clear pattern of high HIV prevalence in Southern Africa and relatively low prevalence in West Africa.
All but two surveys found a higher HIV prevalence among urban residents compared to rural (table 3). While only one survey had an urban:rural ratio below one, the 95% confidence interval included 1 for 5 out of 18 surveys. The urban:rural prevalence ratio varied from 0.95 in South Africa and 1 in Senegal to 3.32 in Rwanda and 3.73 in neighbouring Burundi. There was much variation in the urban:rural ratio in West‐Africa (ranging from 1.0 in Senegal to 3.23 in Niger). The urban:rural ratio appears higher in East Africa (1.65, 1.79, and 2.06 in Uganda, Kenya, and Tanzania respectively) than in the southernmost countries in Southern Africa (0.95 and 1.13 in the two surveys in South Africa, 1.21 in Zimbabwe, 1.33 in Lesotho). The median urban:rural ratio across all eligible surveys was 1.66 (interquartile range 1.14–2.27), while the weighted median urban:rural ratio was 1.65 (interquartile range 1.15–2.06).
Table 3HIV prevalence in urban and rural areas
All but one of the surveys found a higher HIV prevalence among women compared to men (table 4). The female:male prevalence ratio varied from 0.95 in Burkina Faso and 1.07 in Sierra Leone to 2.0 in Zimbabwe (age range 15–29), 2.11 in Guinea and 2.25 in Senegal. The 95% confidence interval included 1 for 7 out of 19 surveys. The median female:male ratio across all eligible surveys was 1.46 (interquartile range 1.24–1.8), while the weighted median female:male ratio was 1.66 (interquartile range 1.37–1.8).
Table 4HIV prevalence among female and males
Table 5 shows the results of scenarios assuming that non‐responders have higher HIV infection levels than those who accepted the HIV test in the survey, with relative risks of 1.25, 1.5, and 2, for countries with sufficient information on the levels of non‐response.
Table 5Scenarios of adult HIV prevalence assuming different risks of prevalence for the non‐tested relative to those who were tested
For countries with high response levels, the overall adjusted prevalence is not very different from the prevalence observed in the survey. For example, in Rwanda with non‐response of less than 4%, even with a relative risk of 2 for non‐responders, the adjusted prevalence would be only 0.1% higher than the observed, with a ratio of adjusted versus observed prevalence of 1.03. However, for countries with significant levels of non‐response, the adjusted prevalence can be very different from the observed. For example, in the South Africa HSRC survey with non‐response of 34%, a relative risk of 2 for non‐responders results in an adjusted prevalence of more than 5% higher (at 21.6%) than the observed prevalence, with a ratio of adjusted versus observed of 1.34.