91��ɫ

As avian influenza spreads across species in new ways, a study with contributions from 91��ɫ Associate Professor Iain Moyles suggests the mathematical models used to track it have not kept pace with reality.

For decades, bird flu, a virus that affects avian species’ respiratory and digestive systems, has been treated as a problem related to poultry farms and migratory birds, says Moyles. In recent years, however, it has spread beyond those boundaries, appearing in mammals, infecting dairy cattle and, in some cases, humans.

That shift requires better understanding of how the virus is evolving and circulating. “Avian influenza needs to be monitored carefully, as its ability to adapt and move between species raises concerns about its outbreak potential,” says Moyles, who teaches in the Faculty of Science's Department of Mathematics and Statistics.

As with other infectious diseases, researchers rely on mathematical models to explore how bird flu might spread in situations that are difficult – or impossible – to observe directly, including rare spillover events where the virus jumps from one species to another.

These frameworks help decision-makers estimate how quickly a virus spreads, which species are most at risk and how effective control measures might be. Their accuracy can directly shape how outbreaks are monitored and managed. While bird flu is not currently transferring easily between people, its growing ability to infect a wider range of animals has raised concerns that it could increase risk for humans.

Moyles and his collaborators, however, noted that as bird flu was identified in new species, including dairy cattle in 2024, there had been little review on whether existing models capture emerging risks.

“This prompted us to perform a modern systematic review of mathematical modelling literature related to avian influenza to identify gaps,” says Moyles about the work now published in .

The research team examined 30 peer‑reviewed studies published between 2023 and mid‑2025 on how bird flu spreads to assess whether current approaches reflect how the virus behaves today. The work was grounded in a “One Health” perspective, which looks at connections between human, animal and environmental health.

What they found is most frameworks rely on a single, dominant approach. Nearly 90 per cent of the studies used compartmental models, which group populations into simple categories – for example, those who can catch the virus, those who are infected and those who have recovered. These models are relatively straightforward to work with and are often used to estimate whether an outbreak is likely to grow or fade.

The problem, researchers say, is many of these models are still based on earlier assumptions about how bird flu spreads, when infections were largely confined to wild birds and poultry. Since 2023, the virus has presented widely in birds, been detected in dozens of mammal species and, notably, has affected dairy cattle. Human infections remain rare, but are increasingly linked to contact with livestock. Despite this shift, most models overlook livestock almost entirely, limiting how well they capture bird flu transmission data and related emerging risks.

Studies using agent-based models, which simulate how individual animals or farms interact, and network models, which map how connections like trade or movement spread disease, was far less common. While these approaches can better capture real-world complexity, such as how poultry movements or farm networks influence transmission, they also require more detailed data and computing power.

The review also found broader gaps in how these models are built and tested. Many rely on assumptions or data from older studies, rather than being grounded in current outbreaks. Important features of the virus – such as how long it incubates, how immunity works or how it persists in the environment – are often simplified. Perhaps most striking, notes Moyles, is that very few models are tested against real-world data. Only two of the 30 studies compared predictions with actual outbreaks, raising questions about how reliable the methods for predicting virus behaviour is.

Models that leave out livestock, oversimplify immunity or fail to incorporate real evidence may give a false sense of certainty, especially when used to guide policy in fast‑moving outbreaks where the virus behaves differently than in the past.

“A lack of understanding of the multi-host transmission and spillover of avian influenza creates gaps in surveillance capabilities,” says Moyles.

To address this, the researchers recommend a shift toward more biologically grounded and data-driven modelling by accounting for livestock, incorporating data on how immunity works, including antibody responses and testing model predictions against real-world outbreaks. They also suggest using multiple approaches together, rather than relying on a single forecast, and call for better surveillance data and greater transparency in how models are built and reported.

“We hope that this review will motivate more One Health mathematical modelling studies and the collection and use of data that support them. This will improve the ability to design meaningful intervention strategies,” says Moyles. “In the long term, the goal is to develop models that better capture spillover risk, including early warning signs of the virus emerging in new species and its potential to cause outbreaks.”

Tags: