Background
Highly Pathogenic Avian Influenza (HPAI) is a prominent candidate for a future human pandemic arising from a zoonotic spillover event. Its best-known subtype is H5N1, with South- or South-East Asia a likely location for an initial outbreak. Such an outbreak would be initiated through a primary event of bird-to-human infection, followed by sustained human-to-human transmission. Early interventions require the extraction, integration and interpretation of epidemiological information from the limited and noisy case data available at outbreak onset.

Methods
We studied the implications of a potential zoonotic spillover of H5N1 influenza into humans. Our simulations used BharatSim, an agent-based model framework designed primarily for the population of India, but which can be tuned easily for others. We considered a synthetic population representing primary contacts in an outbreak site with infected birds. These primary contacts transfer infections to secondary (household) contacts, from where the infection spreads further. We simulate outbreak scenarios in farm as well as wet-market settings, accounting for the network structure of human contacts and the stochasticity of the infection process. We further simulated multiple interventions, including bird-culling, quarantines, and vaccinations.

Results
We show how limited, noisy data for primary and secondary infections can be used to estimate epidemiological transmission parameters, such as the basic reproductive ratio R_0 from other metrics like the secondary attack risk, in realistic social interaction settings. We describe the impact of early interventions (bird-culling, quarantines, and vaccination), taken together or separately, in slowing or terminating the outbreak.

Conclusions
An individual-based model allows for the most granular description of the bird-human spillover and subsequent human-to-human transmission for the case of H5N1. Such models can be contextualised to individual communities across varied geographies, given representative contact networks. We show how such models allow for the systematic real-time exploration of policy measures that could constrain disease-spread, as well as guide a better understanding of disease epidemiology for an emerging infectious disease.