BAYESIAN PROJECTION OF THE ACQUIRED-IMMUNE-DEFICIENCY-SYNDROME EPIDEMIC

Short-term projections of the acquired immune deficiency syndrome (AID S) epidemic in England and Wales have been regularly updated since the publication of the Cox report in 1988. The key approach for those upd ates has been the back-calculation method, which has been informally a dapted to acknowledge various sources of uncertainty as well as to inc orporate increasingly available information on the spread of the human immunodeficiency Virus (HIV) in the population. We propose a Bayesian formulation of the back-calculation method which allows a formal trea tment of uncertainty and the inclusion of extra information, within a single coherent composite model. Estimation of the variably dimensione d model is carried out by using reversible jump Markov chain Monte Car lo methods. Application of the model to data for homosexual and bisexu al males in England and Wales is presented, and the role of the variou s sources of information and model assumptions is appraised. Our resul ts show a massive peak in HIV infections around 1983 and suggest that the incidence of AIDS has now reached a plateau, although there is sti ll substantial uncertainty about the future.