SummaryThis article extends existing methods for estimating reporting delays to allow nonparametric estimation of how delays are changing over time. Also implemented are refinements to estimate calendar month effects and to improve the accuracy of trend estimates by focusing on delays of >6 months. Applying the method to 1987-definition adult and adolescent AIDS cases reported by June 1992 shows strong evidence for a nonlinear trend toward longer delays among cases diagnosed more recently and for slower reporting of cases diagnosed in January and June of each year. Combining estimated reporting delay corrections with the possibility of increasing underreporting produces a 14–16% higher estimated incidence by December 1991 and a 19–24% higher projected incidence by December 1993 than using the delay corrections provided with the public information AIDS data and assuming constant underreporting rates.