The Principal Investigator proposes to continue a research program in modeling and decision support pertaining to the AIDS epidemic. Special features such as periodic group encounters and migration from group to group suggest that the epidemic of infection with HIV among IVDU's differs sufficiently from the well-studied homosexual epidemic as to require distinctive mathematical representation. Models developed under the current grant are natural for the IVDU epidemic, but impose analytic and computational challenges. In this project, computational methodology based on generalizing the binomial and multinomial chain approach will be used to meet the computational challenge of IVDU/HIV analysis. Methods planned for assessing approximation accuracy may well be useful also for conventional AIDS models already in the literature. Unsettled issues remain about the adequacy of conventional deterministic approximations. The objective of AIDS modelling ought ultimately to be a framework for intervention and control of the disease. A substantial effort will be given to placing dynamic models (as above) into a quantitative decision framework by assessing costs of alternative intervention policies and examining the resulting decision problems. The challenge is considerable, because no precedent exists for the evolution of the process. Special attention will therefore be placed on """"""""automatic learning"""""""" techniques which require very few process assumptions yet, nevertheless, assure asymptotic optimality. A practical focus for this research is the Pima County (AZ) IVDU epidemic and the provision of suggestions to prevent it from following the prevalence trends of larger metropolitan areas. Toward that end, ties already made with local health officials and epidemiologists at San Francisco General Hospital will be strengthened.
| Yakowitz, S (1995) Computational methods for Markov series with large state spaces, with application to AIDS modeling. Math Biosci 127:99-121 |
