Demographic and Spatiotemporal Patterns of Avian Influenza Infection at the Continental Scale, and in Relation to Annual Life Cycle of a Migratory Host

Abstract

Since the spread of highly pathogenic avian influenza (HPAI) H5N1 in the eastern hemisphere, numerous surveillance programs and studies have been undertaken to detect the occurrence, distribution, or spread of avian influenza viruses (AIV) in wild bird populations worldwide. To identify demographic determinants and spatiotemporal patterns of AIV infection in long distance migratory waterfowl in North America, we fitted generalized linear models with binominal distribution to analyze results from 13,574 blue-winged teal (Anas discors, BWTE) sampled in 2007 to 2010 year round during AIV surveillance programs in Canada and the United States. Our analyses revealed that during late summer staging (July-August) and fall migration (September-October), hatch year (HY) birds were more likely to be infected than after hatch year (AHY) birds, however there was no difference between age categories for the remainder of the year (winter, spring migration, and breeding period), likely due to maturing immune systems and newly acquired immunity of HY birds. Probability of infection increased non-linearly with latitude, and was highest in late summer prior to fall migration when densities of birds and the proportion of susceptible HY birds in the population are highest. Birds in the Central and Mississippi flyways were more likely to be infected compared to those in the Atlantic flyway. Seasonal cycles and spatial variation of AIV infection were largely driven by the dynamics of AIV infection in HY birds, which had more prominent cycles and spatial variation in infection compared to AHY birds. Our results demonstrate demographic as well as seasonal, latitudinal and flyway trends across Canada and the US, while illustrating the importance of migratory host life cycle and age in driving cyclical patterns of prevalence.

Fig 1. Locations of capture sites of Blue-winged Teal sampled for low pathogenic avian influenza virus across Canada and the United States, from 2007 to 2010.

Fig 2. Annual and seasonal means of predicted probability of avian influenza virus infection in Blue-winged Teal in Canada and the US from 2007–2010.

Point predictions were based on the best-supported model (Table 3) and were averaged across both sexes, and all flyways and latitudes of data. Confidence intervals are calculated based on the variance around the group-mean within each year-season-age category. Hatch year (HY) and after hatch year (AHY) age groups shown separately.

Fig 3. Predicted probability of avian influenza virus (AIV) infection in Blue-winged Teal, illustrating the interaction between age and season.

Predictions are based on the best-supported model (Table 3) with other explanatory variables set at male (Sex), 2010 (Year), central flyway (Flyway) and mean latitude. Hatch year: HY, after hatch year: AHY. Error bars represent 95% confidence intervals, based on SE calculated by the delta method.

Fig 4. Probability of avian influenza virus infection in Blue-winged Teal as a function of latitude of sampling location.

Predicted probability was calculated based on the best-supported model (Table 3) with Age, Sex, Flyway, Season and Year set at hatch year, male, central, July-August and 2010 categories, respectively. Error bars represent 95% confidence intervals.

Fig 5. Predicted risk of avian influenza virus infection in hatch year Blue-winged Teal in Canada and the US from 2007–2010 at different stages of the annual life cycle.

Natural neighbor interpolation spatial analysis was applied to predicted probability values calculated for all hatch year birds based on the best-supported model (Table 3), and averaged across all years and both sexes for each sampling site (circles).

Fig 6. Predicted risk of avian influenza virus infection in after hatch year Blue-winged Teal in Canada and the US from 2007–2010 at different stages of the annual life cycle.

Natural neighbor interpolation spatial analysis was applied to predicted probability values calculated for all after hatch year birds based on the best-supported model (Table 3), and averaged across all years and both sexes for each sampling site (circles).