Modeling seasonal variation in rotavirus hospitalizations for use in evaluating the effect of rotavirus vaccine.

Every year rotavirus epidemic repeats in cooler months of the year in temperate countries, but the size of the epidemic may often vary. Such seasonal variation needs to be considered when the effect of rotavirus vaccine is predicted before vaccine introduction or it is evaluated after vaccine introduction. A computer program based on a stochastic decision tree model was developed to produce stochastic variation, which was used as a proxy for seasonal variation, in the number of rotavirus hospitalizations. When the model was applied to a hypothetical community with a birth cohort of 1,000 children in Japan, it predicted the occurrence of up to 29% of stochastic variation from the average number of rotavirus hospitalizations. Then, the model was applied for use in evaluating the effect of rotavirus vaccine in two different scenarios regarding vaccine use in the community: a scenario where rotavirus vaccine was introduced only into the private sector, and another where it was incorporated into the universal immunization program. In the former scenario, an average of 23% reduction in the number of rotavirus hospitalizations was predicted, but this level of reduction would be obscured due to seasonal variation. In the latter scenario, an average of 74% reduction was predicted, which would be beyond seasonal variation. This model will be useful to inform stakeholders and policymakers how vaccine introduction will change the burden of rotavirus disease under different scenarios of vaccine implementation. (c) 2010 Wiley-Liss, Inc.