| Summary: | For monitoring the spread of HIV epidemic, both national population-based surveys and antenatal clinics (ANC) are
used. However, in all cases, there are potential biases. Bias associated with ANC data includes whether the pregnant
women who attend public ANC are representative of all pregnant women. Reduced fertility among HIV-infected
women, selection for sexual activity and under-representation of smaller rural sites in surveillance systems are other
factors that may be source of biases (Boerma et al. 2003 & Walker et al. 2003). So, the question arising is how
women who attend ANC could be representative of the general female population. Evidently, not all women become
pregnant and not all pregnant women attend ANC.
This research project has been designed to address those biases especially in Rwanda and Malawi. It focused on
investigating the significance of this bias by doing a comparative analysis of sero prevalence between both those
using ANC and those who do not. This study, therefore, intends to test whether women attending ANC may be
representative to the general female population of both Rwanda and Malawi using respectively 2004 MDHS and
2005 RDHS.
Using statistical techniques with the aid of STATA software program, univariate, bivariate and logistic regression
(bivariate and multivariate) were performed for 11321 women in Rwanda and 11698 in Malawi aged between 15
and 49. However, among them, those who had live birth in last five years prior to the surveys were the most
interested on in this study; that is especially, 5390 in Rwanda and 7304 in Malawi. Besides, HIV status of
respondents was an important variable.
Considering both women who had live birth and those who did not have live birth, I find that women who had live
birth in Rwanda are 0.62 times less likely to be HIV positive and 0.48 times less likely to be infected for those who
had live birth in Malawi. When controlling for women who had live birth, I find that in both countries women who
use ANC are less likely to be infected compared to those who do not (0.53 times less likely in Rwanda and 0.58
times less likely in Malawi).
Based on these findings, relying only on data from ANC may lead to biases in HIV prevalence estimates;
particularly referring to 2004 MDHS and 2005RDHS. Besides, considering the level of significance of the
difference between HIV status between those who use ANC and those who do not, I find that this is not identical in
Rwanda (5% level of significance) and in Malawi (10% level of significance). Thus, these results suggest, briefly,
that not only the degree of ANC data representativeness is changing depending on various stages of HIV epidemic
as Fylkesnes said (1998), but also is affected by the amount of women who had live birth and their respective HIV
status. In fact, this difference may be based on the fact that in Malawi, HIV prevalence is high compared to Rwanda
and those who had live birth were in high percentage comparing to Rwanda.
|
|---|
