Daniel Merl and Marc Mangel
12/31/2005 09:00 AM
Applied Mathematics & Statistics
Increased national concern over the possibility of sudden and wide-spread outbreak of infectious disease motivates the modeling of pathogen transmission for various characterizations of the underlying population structure. A critical, but often overlooked, component to preparation of this sort is the development of a methodology for defining optimal treatment policies to be deployed in the event of such an outbreak. In realistic scenarios, such a policy must weigh the costs of available treatments (e.g. vaccination or quarantine) against the current state of the epidemic, in terms of numbers of susceptible and infected individuals, and current estimates of the parameters of the transmission model. We present a decision-theoretic framework for adaptive management of epidemiological interventions. We derive an expression for the optimal adaptive policy conditional upon current information about the epidemic, evaluation of which is computed using stochastic dynamic programming. We present simulation studies to demonstrate the advantages, in terms of reduced expected loss and more efficient resource management, of the optimal adaptive policy over the optimal non-adaptive policy.
