This study evaluates the overall impact of the bias due to HIV/AIDS on the own-child method of estimating child mortality, and the contribution of the various components to the overall bias. Indirect estimates of child mortality are calculated by applying the own-child method to the 2009 Lesotho Demographic and Health Survey (LDHS) and the 2005-6 Zimbabwe DHS. In the process, the potential of the software, MATCHTAB, to link successfully surviving children to their biological mothers is demonstrated using the 2006 Lesotho census data. To evaluate the impact of HIV/AIDS on the own-child method, the target estimates of child mortality for Lesotho and Zimbabwe are also developed by adjusting the direct estimates from the full history data for the selection has introduced by HIV/AIDS.
While the overall bias due to HIV/AIDS induces a significant underestimation of child mortality, its impact appears to be diminished by the upward bias inherent in the indirect own-child method relative to the direct method. As expected the magnitude of the overall bias is highest for the estimates derived from the reports of women aged 30-34, the oldest age group used in the study. The results for Zimbabwe are also consistent with findings from other studies indicating that the bias due to HIV/AIDS is greater for the under-five than the infant mortality rate. However, for Lesotho, it is difficult to infer the differential impact of the overall bias on the infant and under-five mortality rate because of the confounding effect of the relative methodological bias.
For both countries, the magnitude of the underestimation of child mortality induced by the absence of data on the survival of children of mothers who have died (selection bias) exceeds the 5 per cent threshold and is thus significant. Further research is needed to develop correction factors that adjust for the selection bias introduced by HIV/AIDS into the own-child method.
Although converting the mortality rates to a common index of mortality using a standard table which does not allow for HIV leads to the underestimation of the underfive mortality rate, the underestimation is not significant. However, there is a need for the development of model life tables that incorporate the impact of the HIV/AIDS epidemic for use in countries with a high prevalence of HIV/AIDS. Given the significant impact of the selection bias, the impact of HIV/AIDS on the own-child method is not addressed adequately by only replacing the non-AIDS Princeton model life tables with standard life tables that incorporate the impact of the HIV/AIDS epidemic.