Influenza A virus epidemiology – from individual disease histories to disease dynamics

5566951033_902c131ed8_b

Mallards on the wing (Photo by Flickr user Bengt Nyman used under a CC-BY 2.0 license)

Wildlife disease studies are challenging. That’s a fact. If you want an easy science life you should choose another path with more instant results. However, challenging is also the opposite of boring, and the rewards of getting your results are even more exhilarating when lots of toil, sweat and tears have been invested. As readers of this blog are aware, wildlife disease studies are what we do, and I have repeatedly written about our ongoing work on influenza A virus ecology and epidemiology in wild migratory Mallards. This week another study from our study site was published, entitled Capturing individual-level parameters of influenza A virus dynamics in wild ducks using multistate models, which can be found on early view in the Journal of Applied Ecology.

The challenges of studying wildlife disease dynamics are that you want to capture a dynamic process influenced both by the host and the pathogen, which in turn is compounded by variation in the environment – both biotic factors, such as food abundance and the occurrence of other potential hosts, and abiotic factors, such as weather and climate. Disentangling these interconnected effects is a little like making a cube out of mercury. In most wildlife disease studies the available data is at the population level, usually in the form of prevalence rates at specific time points. This type of data is ‘fairly easy’ to collect – you head out into the field, sample all animals you can lay your hands on and then use this snapshot in time as a proxy for the true disease dynamic in your system. The more times you are out collecting data, the better your model becomes. However, disease is driven by factors operating at the level of individuals, such as infection risk and recovery rate, and that type of data can only be acquired by repeated sampling of individuals across a suitable timescale. This is rarely achieved because of logistical, practical and monetary reasons.

26383076326_62d7d29903_k

Mallards on the wing (Photo by Flickr user Bengt Nyman used under a CC-BY 2.0 license)

We, however, sit on a huge collection of Mallard and flu data gathered at the same study site with similar methods over a period of close to 15 years. Our latest paper, headed by Alexis Avril and with collaboration with colleagues in France, utilizes this dataset to develop individual-based influenza A virus epidemiological models. This proved to a monumental task that stretched over several years and burned the processors of a good number of computers. Part of the difficulty can be attributed to the data itself – capture and disease histories for 3500 individuals collected over 7 seasons, where at each capture occasion axillary data on bird age, sex, condition, infection status and weather were included. But also the patchy nature of recapture probability and the short duration of most influenza virus infections contributed significantly to extensive data crunching.

Picture2

The conceptual framework in the multistate CMR model.

The method we used was multistate capture-mark-recapture models, which are extensions of models originally developed to investigate mortality rates from census data, but where one can include the infection state – i.e. infected or not with influenza virus – as a factor in the analyses. Interested readers should head over and read the publication, as I will spear the rest of you any hardcore statistics and model lingo. Parts of the abstract serves as a good summary:

 For most years, prevalence and risk of influenza A virus (IAV) infection peaked at a single time during the autumn migration season, but the timing, shape and intensity of the infection curve showed strong annual heterogeneity. In contrast, the seasonal pattern of recovery rate only varied in intensity across years. Adults and juveniles displayed similar seasonal patterns of infection and recovery each year. However, compared to adults, juveniles experienced twice the risk of becoming infected, whereas recovery rates were similar across age categories. Finally, we did not find evidence that infection influenced the timing of emigration from the stopover site.

Our study provides robust empirical estimates of epidemiological parameters for predicting IAV dynamics. However, the strong annual variation in infection curves makes forecasting difficult. Prevalence data can provide reliable surveillance indicators as long as they catch the variation in infection risk. However, individual-based monitoring of infection is required to verify this assumption in areas where surveillance occurs. In this context, monitoring of captive sentinel birds kept in close contact with wild birds is useful. The fact that infection does not impact the timing of migration underpins the potential for mallards to spread viruses rapidly over large geographical scales.

Our findings corroborate much of the earlier works done on IAV in birds from population level data or from infection experiments, but with higher robustness of the conclusions. Importantly, we provide estimates of the most crucial infection parameters and show how they vary in relation to age in different seasons and years. And from a model point of view, we show that MS-CMRs are a potent method for disease dynamic inferences. We hope this paper will be read and cited by people in the IAV field and in general disease dynamic research, and that it will be useful for stakeholders interested in the contribution of wild birds in the epidemiology of IAV in poultry.

Link to the paper:

Avril, A., Grosbois, V., Latorre-Margalef, N., Gaidet, N., Tolf, C., Olsen, B. & Waldenström, J. 2016. Capturing individual-level parameters of influenza A virus dynamics in wild ducks using multistate models. Journal of Applied Ecology, online early.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s