Introduction We examined the epidemiological diversity on the incidence of H1N1 within and among small regional communities using surveillance data and agent-based simulations.

Methods We investigated 27 elementary and junior high schools in Moroyama-town and Sakado-city located in the central part of Saitama Prefecture, Japan. The surveillance system was built on a www server. Agent-based modelling and simulations were performed using AnyLogic 6.5.1 (XJ Technologies, St.Petersburg).

Results By the end of March 2010, cumulative incidence rate (CIR) of H1N1 among school children reached 30% and 34% in Moroyama and Sakado, respectively. There was no considerable difference between epidemic curves in these neighbouring town and city. On the other hand, in the individual schools, the CIRs ranged 16%–51% even if the schools are closely located. To examine the cause of this diversity, we performed agent-based modelling and simulations assuming inequal probability of infection within and outside of schools. Repetitive simulations gave CIRs of 23%–44%, indicating that the CIRs of the small population communities may considerably vary even though all the agents were assumed to have the same susceptibility to infection.

Conclusion The granularity of surveillance/analyses/prevention should be finer than in the past to achieve the most effective policies against influenza and similar communicable diseases in the local communities. The cause of this diversity can be explained in part by the stochastic nature of infection transmission processes in the small populations shown by the agent-based simulations. Relevance of the other issues, for example, environmental factors, vaccination, intrafamilial infection, etc, is currently under investigation.