The effect of HIV on the orphanhood method of estimating adult female mortality

Includes bibliographical references (leaves 73-74).The orphanhood method of estimating adult mortality is widely used in developing countries. The method is subject to a number of assumptions, some of which are violated when a generalized high-prevalence HIV epidemic is present. Non-independence of the mortality of children and mothers, relationships between HIV infection and fertility, and changes in age-specific mortality result in biases that affect the accuracy of the method. An earlier study has examined some of these sources of error, and proposed adjustments to enable continued use of the method. This earlier research, however, uses data from populations with lower HIV prevalence rates than are currently being experienced in much of southern Africa, and is based on specific assumptions about HIV and its effects on mortality and fertility. The effects of HIV on the method are investigated in this research using mathematical modelling of the effects on Black South African females - a population with high HIV prevalence. More is now known about HIV and its effects on mortality and fertility, and these effects are explicitly reflected in the ASSA2002 model which provides much of the data for this research. The research compares the simulated survival of various cohorts of women: those aged 25 in a certain year, women (with an age profile identical to that of mothers), mothers, and mothers as reported by their children. In this way the various sources of error are explicitly identified and the errors quantified. The timing, magnitude, and combined effects of the errors are studied in relation to the emergence and spread of HIV, indicating when the errors might be expected to be large enough to invalidate the method. Errors that bias the outcomes of the orphanhood method take a number of years to develop after HIV starts spreading. Substantial biases in reported survival emerge between 20 and 35 years after the start of an HIV epidemic, in a high prevalence setting. These errors are reduced by the use of antiretroviral and prevention of mother-to-child transmission, but biases remain large enough to invalidate outcomes when the unadjusted method is applied in most southern African countries. An adjusted method has been proposed which substantially reduces error, except when adjusting survival reported by the two youngest age groups. This adjusted method can be applied, but further research to identify revised adjustments would further improve the accuracy of the method