Limitations of the possible model

One of the limitations of the Demographic Impact of HIV/AIDS in Mozambique concerns the fact that it was drawn up on the basis of seroprevalence data obtained from just five urban sites in the centre and south of the country in 1998. The study of the impact of HIV/AIDS on the education sector was also based on these projections. The data presented here are based on the Spectrum projection models. This model, developed jointly by the INE and the Health Ministry to assess the demographic impact of HIV/AIDS in Mozambique, drew up projections of the population up to 2010 based on the Rural and Urban Projection programme (RUP). The spectrum system of policy models was then applied to project the impact of HIV/AIDS. Two sub-routines within Spectrum were used - DemProj and AIM (AIDS Impact Model). DemProj is the demographic model within Spectrum and is used to draw up population projections. AIM is the model used to assess the impact of the HIV/AIDS epidemic. AIM uses various assumptions about the population, fertility, mortality etc. One cannot stress too much that models are, by definition, representations of aspects of reality, and are as solid as the theories on which they are based and the reliability of the data used. Consequently, the projections resulting from these models are limited by the quality of the available data, something which is a particular problem in the situation of Mozambique. In Mozambique, there have been problems of information associated with the data of the 1997 census, and the data from the observation at the sentinel sites. The model took as its population base year a synthetic recreation of the 1981 population, based on the results and assumptions of the 1997 Census on levels of fertility and mortality prior to 1997.

The observation data (used to estimate prevalence rates among adults) were only available for five urban centres. The technicians started from the assumption that the epidemic began to spread earlier in the centre of the country than in the south. Furthermore, the assumption in the model was that seroprevalence in the centre was double that in the south, and that by 2005 it would be 50% higher. Since there was no observation in the three northern provinces, it was necessary to establish a basis for estimating HIV prevalence among the adult population of that part of Mozambique. The model assumes that in 1998 the prevalence was 25% higher than in the south, and that the prevalence rates in southern and northern Mozambique will be the same in 2005. Data on the epidemic will certainly benefit from expanding the number of sentinel sites from 4 to 22, distributed as shown in Map 4.1.