The Enigma of HIV/AIDS Statistics in Africa
Population-based surveys and antenatal data put to question
Leela McCullough recently alerted the AFRO-NETS group of the
preliminary findings of the Kenyan DHS 2003 which reveals a
lower estimate of HIV/AIDS prevalence in Kenya than previously
thought. In this mail, I highlight this discrepancy.
The Enigma of HIV/AIDS Statistics in Africa: Population-based
surveys and antenatal data put to question
Accurate information about disease occurrence is needed for
proper monitoring and management. This is true in both condi-
tions affecting just a few as well as more epidemic conditions
that affect in pandemic dimensions like HIV/AIDS. Unfortunately,
current estimates of HIV/AIDS occurrence in several African
countries are highly enigmatic. Statistics are starkly incoher-
ent and a plethora of articles in lay-publications (News-papers
etc.) are quickly pointing out this gap.
In the wake of the release of Kenya?s DHS 2003 preliminary re-
port for example, one of South Africa?s leading papers was quick
to publish a lead article in it?s cover page titled: "HIV/AIDS
figures are overestimated". Interestingly, a number of articles
(on this topic) had been published earlier by the same paper but
the majority of them had been less conspicuously situated in the
inside pages of the newspaper. It will however appear that the
Kenyan Study gave impetus to the controversies around HIV/AIDS
statistics that the topic qualified to assume cover page. But
what really is this about and how can it be explained.
The Kenya DHS incorporated HIV prevalence survey into it's pro-
Leela McCullough alerted the AFRO-NETS group on this and re-
ferred us to http://www.afronets.org/pubview.php/93 which con-
tained the preliminary report of Kenya?s DHS. In summary, the
DHS indicates that the estimated prevalence of HIV/AIDS in Kenya
based on ante-natal data was an over-estimation when compared
with what was obtained from the DHS population based survey -
9.4% from antenatal survey compared to 6.7% from the population
This corroborates finding in South Africa where a similar popu-
lation based survey has been conducted. The HSRC/Nelson Mandela
Foundation in 2002 reported 11.4% prevalence nationally compared
to estimates of 14.2% (sometimes higher) based on antenatal sur-
veys. Even though all the statistics are in essence "estimates"
and must not claim 100% accuracy, it is important that we have
an idea of the magnitude of variation of estimates from the "ac-
tual values" and why. Not doing this will convey a wrong message
to the public at the detriment of their health and well-being.
Grossly higher estimates wrongly cause tension to the health
system at the expense of other (ACTUAL) problems. Grossly lower
estimates will present a wrong judgment of a massive public
health problem that undermines its magnitude. Incoherent esti-
mates raise skepticism on the truth of the existence of a prob-
lem (big or small)! It is therefore important that Epidemiolo-
gists, Demographers, Actuarial Scientists, Biostatisticians and
others involved in the Measurement of Health make a concerted
effort at obtaining "best estimates" which are coherent and in-
terpret this to the lay public.
From the outset, it is important to state that observations of
lower values from population surveys compared to antenatal sero-
prevalence data are to be expected. This is so because antenatal
data makes estimations based exclusively on a high-risk group of
currently or recently sexually active pregnant women who do not
adopt HIV prevention strategies (e.g. condom use). Making gener-
alizations for the rest of the population may therefore overes-
timate the situation in the general population. However, antena-
tal seroprevalence surveys remain an easy, less expensive method
of following up the HIV prevalence in many countries. Indeed,
rising prevalence in this group (pregnant women) will most defi-
nitely imply rising prevalence in the general population.
Population based surveys on the other hand will tend to report
lower values. Assuming there is no selection bias and the sample
is representative enough, values emanating from these surveys
should be more reflective of the actual situation. The draw back
is the cost implication of undertaking such surveys and diffi-
culty in obtaining a sample representative enough.
With these considerations in mind, it is important that estima-
tors of HIV/AIDS statistics be more precise in defining what
they measure. The difficulties associated with obtaining defi-
nite statistics on HIV prevalence need to be explored and novel
ways of synthesizing emerging statistics attempted.
In conclusion, whilst others may dismiss a pursuit for accurate
estimates as "unnecessary" on grounds that "it does not make the
epidemic change", the fallacy of such a notion is echoed in it?s
own statement?."it sure makes the epidemic CHANGE!"
Louis Trichardt, South Africa