Report on the Global HIV/AIDS epidemic

Annex 1.How reliable are estimates of HIV infection and AIDS deaths?
ies in urban and rural communities have made it possible to validate the results of
sentinel surveillance in pregnant women. These studies offer important insights into
the differentials of HIV infection by age and sex. They also confirm that sentinel surveillance in pregnant women yields remarkably robust estimates for HIV prevalence
in the general population of reproductive age, as shown in Figure 28.
Figure 28. HIV prevalence rates among pregnant women and among
all adults aged 15-49
% 35
0 pregnant women
0 all adults
20
15
10
5
Lusaka, Mposhi Mwanza, Rakai, Rakai, Rakai, Kisumu,
Zambia Zambia Tanzania Uganda Uganda Uganda Kenya
1994-1996 1994-1996 1990-1991 1990 1991 1992 1998
Source: Sentinel surveillance data from antenatal clinics and population-based studies, selected African
countries, 1990-1998
In countries where the HIV epidemic is concentrated in a few groups with high-risk
behaviour - mostly drug injectors and men who have sex with men, as well as sex
workers and their clients - the methods for estimating HIV prevalence are different.
This is the case in most countries in Asia, the Americas and Europe. Here, because
there are very few HIV infections in the general population, and because many of the
infections are in groups largely or entirely made up of men, data from pregnant
women are of very limited use. Rather, HIV estimates in such cases are based on
information on HIV prevalence in each group of people with high-risk behaviour,
together with estimates of the size of each of these populations. Since these behaviours are often socially unacceptable and sometimes illegal, information on both HIV
prevalence levels and the size of the population affected can be much harder to
come by. Consequently, uncertainties around these estimates may well be greater
for countries where the epidemic is concentrated in specific groups.

117