Sensitivity analyses for the backcalculation method of AIDS projections.

In Australia, AIDS is a notifiable disease and the data have been of a relatively high quality. In this article, the cases among Australian residents diagnosed by mid-1989 and reported by January 1990 are used to evaluate the sensitivity of estimates of numbers infected with the human immunodeficiency virus (HIV) and of projections of numbers of new AIDS diagnoses to major uncertainties in the method of backcalculation. We find that the estimates and long-term projections are sensitive to all uncertainties, including the incubation period distribution, the new infection intensity distribution, and the level of data aggregation chosen for analysis. Further, we find the projection that is least sensitive to all uncertainties, namely AIDS diagnoses for the next year (mid-1989 to mid-1990), falls in a relatively narrow range that encompasses the number that had been reported by April 1991.