Identifying spatio-temporal patterns of transboundary disease spread: examples using avian influenza H5N1 outbreaks

Characterizing spatio-temporal patterns among epidemics in which the mechanism of spread is uncertain is important for generating disease spread hypotheses, which may in turn inform disease control and prevention strategies. Using a dataset representing three phases of highly pathogenic avian influenza H5N1 outbreaks in village poultry in Romania, 2005-2006, spatiotemporal patterns were characterized. We first fit a set of hierarchical Bayesian models that quantified changes in the spatio-temporal relative risk for each of the 23 affected counties. We then modeled spatial synchrony in each of the three epidemic phases using non-parametric covariance functions and Thin Plate Spline regression models. We found clear differences in the spatio-temporal patterns among the epidemic phases (local versus regional correlated processes), which may indicate differing spread mechanisms (for example wild bird versus human-mediated). Elucidating these patterns allowed us to postulate that a shift in the primary mechanism of disease spread may have taken place between the second and third phases of this epidemic. Information generated by such analyses could assist affected countries in determining the most appropriate control programs to implement, and to allocate appropriate resources to preventing contact between domestic poultry and wild birds versus enforcing bans on poultry movements and quarantine. The methods used in this study could be applied in many different situations to analyze transboundary disease data in which only location and time of occurrence data are reported.