Multiple imputation method for estimating incidence of HIV infection. The Multicenter Prospective HIV Study.

Background. CD4+ T-lymphocyte (CD4) and platelet counts are good predictors of the 'maturity' of HIV infection and can be used to impute the date of infection/seroconversion in individuals for whom this date is unknown. Methods. Data from the Italian Seroconversion Study were used to develop a Weibull regression model for time since seroconversion as a function of the haematologic markers. The model was used to impute time since HIV infection/ seroconversion in individuals from a prevalent cohort, recruited through the Lazio regional HIV surveillance system. Results. The range of the imputed calendar times of infection/seroconversion in 2599 HIV prevalent individuals was 1972-1992; the earliest seroconversions occurred among injecting drug users (IDU). The peak of incidence was reached in 1986 with 340 seroconversions. Among males, the estimated median time from seroconversion to HIV diagnosis was shorter in IDU (30 months) as compared to non-IDU (36 months). This difference was smaller for females (26.6 versus 28.4 in IDU and non-IDU, respectively). Conclusions. This method permits the estimation of population-based curves of HIV incidence, using data from surveillance. The results support the hypotheses of an early spread of the epidemic among IDU in the Lazio region, and of shorter lead times in this population.