Measurement of adult mortality in populations affected by AIDS: an assessment of the orphanhood method

This paper demonstrates that orphanhood data can be used to estimate adult women’s mortality in populations experiencing an epidemic of AIDS. It develops both a correction for selection bias in reports of orphanhood and a revised procedure for estimating life table survivorship for use in populations with significant AIDS mortality. These new methods yield mortality estimates for a Ugandan population that are consistent with those obtained by prospective surveillance. Countries that lack effective death registration systems should ask about the survival of mothers in the census and surveys in order to monitor the effect of the AIDS epidemic on mortality. Most developing countries lack effective systems of death registration. Measuring adult mortality in such populations has always been difficult (Timaeus 1991a). The large-scale AIDS epidemics that have developed in much of Africa and some other countries have emphasized the public health importance of monitoring adult mortality levels and trends in the developing world. They have also made this task even more of a challenge. This paper commences by reviewing briefly both the existing limitations of the methods used to estimate adult mortality in the developing world and the additional obstacles to the production of such estimates that arise in populations with substantial mortality from AIDS. All mainland sub-Saharan African countries and most other developing countries lack complete and accurate civil registration systems. In addition, most deaths occur outside hospital. Thus, no possibility exists of compiling comprehensive data on deaths by age routinely and continuously. A straightforwar d and relatively cheap alternative way of generating information on mortality is to ask questions in a national census or single-round household survey about deaths in the past year. Unfortunately, such questions have proved unreliable and often yield incomplete data (Timaeus 1991a). Respondents may find it difficult to recall exactly when a death occurred and often have little idea of the ages of those who have died. Moreover, only very large surveys can estimate adult death rates with reasonable precision. The deaths of adults who lived alone are unlikely to be reported. Furthermore, not everyone is attached clearly to a single household. Thus, both omission of deaths and doublecounting occur. Such ambiguities in the scope of questions about recent deaths in the household are particularly severe when a household head dies, as this event is likely to stimulate the fission and reformation of households. * This study was initiated while Andrew Nunn was working with the Medical Research Council program on AIDS in Uganda. We are grateful to the directors of the program, Dr. D. Mulder and subsequently Dr. J. Whitworth, for facilitating this research and to all the staff of the program who were involved in the collection and processing of data used here.