Historical assessment of some specific methods for projecting the AIDS epidemic.

Extrapolation and back-calculation methods for predicting future acquired immunodeficiency syndrome (AIDS) incidence are assessed by examining how they would have performed at various points in the past. The methods use monthly AIDS diagnosis counts through December 1991 for all adults and adolescents in the United States to predict diagnoses 1-3 years into the future. When used on seasonally adjusted data, the average errors in projections by the different methods are smaller than the possible errors due to uncertainty in AIDS counts. This suggests that measuring and improving the completeness of AIDS surveillance may have more potential to improve projections than methodological enhancements.