Comparing epidemic forecasts for European countries obtained from AIDS incidence and HIV prevalence data

Abstract

Previously, the diffusion of HIV/AIDS among the countries of Europe has been reconstructed using a multiregion epidemic modelling system (Thomas 2000; Smith and Thomas 2001). This model was calibrated to WHO AIDS incidence series for each country which, in most instances, were replicated with a high degree of fit. The size of the epidemic implied by this procedure, however, did not conform to point estimates of cumulative HIV prevalence which indicate an eventual size at least three times greater than that attributable to recorded AIDS cases. This paper, therefore, investigates the consequences of this discrepancy for the space–time structure that is estimated for the epidemic in Europe. The first part considers how the modelling system is adapted to take account of national HIV statistics. To this end, an optimisation method is described for identifying disease transmission rates that are consistent with current cumulative HIV prevalence in each country. This method generates a range of feasible solutions where the eventual size of the epidemic is inversely related to this infectivity rate. The second part derives multiregion simulations for European countries using parameter sets calibrated according to both AIDS incidence and HIV prevalence. The outputs for each set include the predicted starting pathway of the epidemic together with a graph recording the direction of the principal infection vector for each country at the forecast time of peak HIV prevalence. For AIDS based scenarios, the diffusion of the epidemic generally flows across Europe from west to east. In contrast, for HIV scenarios at peak prevalence the direction of flow is partially reversed from east to west. The discussion considers the implications of these findings for disease control in the light of issues concerning the recording of HIV and AIDS statistics.