Background
There is a strong link between influenza in birds and human health because influenza epidemics in human populations occur when viruses that typically inhabit the avian gastrointestinal tract mutate or reassort, enabling them to cross the species barrier to infect people [
1]. Mutations arise in avian influenza virus (hereafter "AIV") due to the high error rate of influenza RNA polymerase and the large population size and short generation time of the virus [
2]. Reassortment is the exchange of RNA segments between distinct influenza viruses. When human influenza viruses and AIV reassort, they produce offspring virions that represent a mixture of the parental types' RNA and are infectious to humans in some cases [
2‐
4]. For example, in 1968, one million people died in an influenza pandemic that resulted from the reassortment of an influenza virus from Ukrainian ducks and a virus that had circulated in people since 1957 [
5‐
8].
Today, outbreaks of H5N1 influenza in Africa, Asia, Europe, and the Middle East further illustrate the human health implications of influenza in birds. (Influenza viruses are classified into "HA" and "NA" subtypes based on surface proteins.) People contract H5N1 by handling infected poultry or wild birds after which the virus binds to receptors in the pulmonary alveoli, causing pneumonia and death due to respiratory failure [
9‐
11]. Since July 2003, there have been 436 human cases of H5N1 in the Eastern hemisphere with a 60% mortality rate [
12‐
14]. In 75% of these cases, the infected people had contact with birds [
10]. However, H5N1 has also evolved limited person-to-person transmission within human families [
7,
15‐
17]. Public health authorities are concerned that the evolution of wider human-to-human transmission could result in a H5N1 pandemic that could cause up to 142 million deaths at a cost of $US 4.4 trillion [
14,
18]. The ongoing human pandemic of H1N1 influenza, which has caused over 296,000 human cases and at least 15,921 deaths since mid-February 2009, contains genes from avian, human, and swine influenza viruses [
12].
To date, influenza viruses have been isolated from 105 species of wild birds representing 26 families [
16]. In birds, the H5 and H7 AIV subtypes have periodically mutated from a low pathogenicity form (hereafter "LPAI"), which is typically asymptomatic in wild birds, to a highly pathogenic form (hereafter "HPAI") that causes mortality rates of up to 100% in chickens [
2,
19‐
21]. (Our analysis did not distinguish between influenza subtypes or differentiate LPAI from HPAI.) HPAI also differs from LPAI in that the former has more amino acids adjacent to the hemagglutinin cleavage site, which allows it to replicate in a broader range of tissues [for details, see [
2,
22]]. Aside from poultry, no HPAI H5N1 has been detected to date in the US, though six LPAI H5N1 viruses have been detected in North America since 2004 [
23]. AIV mutated from LP to HP form in poultry in the US in the 1920s, in 1984, and in 2004 [
4,
24‐
29]. Although none of these US outbreaks resulted in the infection of humans with HPAI, it is plausible that HPAI could reassort or mutate to become transmissible to people. As few as five amino acid changes can transform the HP influenza virus into an airborne form that is infectious to mammals [
14]. In the event of an HPAI epizootic in migratory birds in the US, these species could spread HPAI across the country along migratory routes because ducks infected with HP H5 remain healthy enough to migrate [
16,
30]. Indeed, HPAI has already been detected in wild birds in Chad, China, Nigeria, and South Africa [
30‐
32].
We analyze the geographic distribution of AIV in wild birds in the US with the goal of inferring where reassortment events might occur and how HPAI might travel if it enters wild bird populations. Our method for detecting the influenza virus in samples from passerines does not determine whether the virus is LPAI or HPAI (see below). However, 67% of our samples are from non-passerines and are known to be LPAI. Thus, this study assumes that most of our AIV-positive samples are LPAI. We model the geographic distribution of AIV to provide insights about how HPAI might spread if it is introduced to the US in the future. Since we cannot guarantee that the passerine samples are LPAI, when referring to samples that tested positive for influenza virus, we will use the term "AIV" rather than "LPAI". As noted in the Discussion, the characterization of the subtype and pathogenicity of AIVs isolated from passerines in the US remains an important area for future research.
Although the monitoring of HPAI viruses is important, another critical issue in AIV biosafety is the detection of H5 and H7 LPAI viruses. LPAI H7 has been transmitted directly to humans in the US in 1976, 2002, and 2003 [
7,
33]. These cases resulted in conjunctivitis, fever, and upper-respiratory tract symptoms of influenza-like illness, but no fatalities. LPAI H5 and H7 can mutate to HPAI relatively easily given the right environment (for example, poultry sheds). We refer the interested reader to Verdugo et al. [
34], who have developed a model for detecting H5 and H7 LPAI in poultry and predicting when they will evolve to HPAI.
The aims of this research are to measure the prevalence of AIV in different species of wild birds in the US and to prioritize geographical regions for future influenza surveillance. Although aquatic birds in the orders Anseriformes and Charadriiformes have been recognized as reservoirs of AIV since the 1970s, much less is known about AIV prevalence in terrestrial birds in the order Passeriformes [
16,
35‐
38]. Examples of Anseriformes (ducks) that have high prevalence of influenza in the US include the Mallard (
Anas platyrhynchos) and the Northern Pintail (
Anas acuta) [
39,
40]. Shorebirds of the order Charadriiformes in which high influenza prevalence has been detected in the US include the Ruddy Turnstone (
Arenaria interpres) and the Red Knot (
Calidris canutus) [
36,
41]. Recent work detected high prevalence of influenza in passerines in China, including the Eurasian Tree Sparrow (
Passer montanus) [
42,
43]. The present study is necessary in order to test the hypothesis that passerines are important reservoirs of AIV in the US. Further motivation for our study comes from the fact that public health agencies have limited funding to test wild birds for AIV. Thus, it is crucial that the establishment of surveillance sites should be as efficient as possible. For example, the number of sites that are monitored should be small but the sites should be located in counties that are most likely to have birds with AIV. We aim to test the hypothesis that environmental variables can be used to predict AIV cases in wild birds. Next, based on the relationship between AIV and environmental predictors, we attempt to identify the US counties most likely to be influenza hotspots for wild birds.
Our study makes the following contributions. First, guidelines formulated by the World Health Organization recognize the importance of epidemiological modelling using tools such as GIS for the control of AIV [
44]. Nevertheless, most previous work on the geographic distribution of AIV has analyzed Asia and Africa [e.g. [
35,
38,
45]]. To date, studies of AIV in wild birds in the US have focused on Alaska [
39,
46,
47]. However, there may be overlooked hotspots of AIV in the contiguous US [
48]. We contribute the first predictions about AIV cases in the contiguous US at the county scale. Second, we analyze new passerine samples from the Atlantic, Mississippi, and Pacific Flyways supplemented with existing samples from online databases to provide the first comprehensive assessment of AIV prevalence in US passerines. The main finding reported in this article is that the prevalence of influenza in passerines is greater than the prevalence in eight other avian orders. The implication of this finding for human health is that, along with poultry and waterfowl, passerines in the US are a potential vector for the transmission of AIV to people [
16,
49].
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
TF designed the study's analytical methods and drafted the manuscript. SS helped design the analytic methods, including the preparation of the satellite data described in the Methods. EC supervised the first year of laboratory activities, helping to design laboratory techniques for influenza testing, and wrote a FileMaker Pro database to summarize the samples. ET supervised the second year of laboratory activities including flu testing and the cataloguing of more than 10,000 cloacal samples. HT helped design the analytic aspects of the study and carried out a review of the literature about flu in wild birds. WB helped design the study's analytic strategy including the selection of the type of remote-sensed images to be used in the spatial model and selected the spatial scale of these images. DD designed the study's field activities including the sampling of passerine birds as part of the MAPS bird banding network. PN reviewed and revised a preliminary draft of the manuscript, provided guidance about the study's analytical approach, and helped design sampling activities as part of the MAPS network. JM supervised field activities as part of the MAWS bird banding network and designed Figure
2 of the manuscript. CJR supervised field activities in the LaMNA bird banding network, which provided flu samples during the migratory period. JA helped supervised the LaMNA bird banding stations and revised a draft of the article. JP directed laboratory activities during both years of flu testing and helped draft the Methods section of the manuscript. TBS helped design the analytic aspect of the study and helped draft the manuscript. All authors read and approved the final manuscript.