Forecasting the number of AIDS cases: an analysis of two techniques

Two methods used to forecast AIDS cases--trend analysis and back calculation--are examined in this paper. I forecast AIDS cases using trend analysis with data on the number of cases reported between January 1984 and March 1990. Forecasts using back calculation methods are based on the assumption that the progression rates from HIV to AIDS slowed in 1988 due to the use of AZT and aerosol pentamidine. With both methods, reasonable models yield widely different forecasts. Analysis also is hampered by a lack of information about the basic determinants of AIDS incidence, changes in the definition of AIDS, underreporting, and uncertainty about the effect of the use of prophylactic drugs. It is not possible to establish confidence limits around forecasts that take into account these sources of biases, and it is important that those who use AIDS forecasts be aware of these uncertainties.