New model shows H5N1 is undetected in US dairy herds

New model shows H5N1 is undetected in US dairy herds

A powerful simulation of H5N1 transmission across 35,974 U.S. herds shows the virus is far more widespread than reported, raising urgent calls for better agricultural surveillance and stronger control of the diseases.

In a recent study published in the journalNature communicationResearchers developed and tested a novel stochastic metapopulation transmission model to provide the scale, key epidemiological data, and highest risk conditions in the ongoing H5N1 influenza epidemic in U.S. dairy cattle. The model simulates H5N1 transmission between 35,974 herds in the United States, with cattle movement informed by probabilistic outputs from the United States Animal Movement Model (USAMM) and verified with Interstate certificates of veterinary inspection data.

Model results predict that West Coast states have the highest disease burden, with Arizona and Wisconsin at the highest risk for future outbreaks. The study highlights gaps in current biosecurity surveillance systems and suggests that dairy outbreaks are in the 2025 forecast, requiring urgent interventions that address these gaps.

background

The United States (US) dairy industry accounts for a significant portion of the country's GDP (3%). For its routine function, the industry requires frequent movement of the 9 million dairy cows. Unfortunately, this practice often contributes to the transmission of communicable diseases (such as avian influenza) between otherwise isolated herds of cattle.

The US dairy industry is currently facing a serious threat – highly pathogenic avian influenza H5N1. The outbreak brought the disease into the spotlight on farms in Texas, Kansas and New Mexico (February 2024). By December 2024, this outbreak had surpassed 720 livestock herd infections and 35 human infections in the United States. Recent phylogenetic research and structural analyzes on the responsible H5N1 strain suggest that a specific single mutation may be enough to allow binding of the human receptor, raising concerns about the country's milk virus reservoir and increasing the risk of virus adaptation to humans.

Unfortunately, there are no research estimates of the size or predictions of the H5N1 epidemic or predictions of future hotspots.

"In previous outbreaks of bovine diseases such as spongiform encephalopathy and oral diseases in the UK, public health responses have been significantly supported by modeling studies to estimate underreporting rates. Estimating key epidemiological mechanisms, and quantifying the impact of control policies."

About the study

The present study addresses these knowledge gaps by designing and developing a stochastic metapopulation transmission model (SEIR) to simulate H5N1 transmission in 9,308,707 dairy cows (35,974 herds) across the continental United States (48 states; 2022 Census data). It uses a Bayesian evidence synthesis approach to estimate epidemiological parameters corresponding to reported outbreaks.

The model simulation was initiated by infecting five cows in Texas based on phylogenetic analyzes suggesting an initial spillover in December 2023, with additional seeding reflecting early outbreaks. Cattle migration between herds was estimated using a likelihood function calculated using data from the United States Animal Movement Model (USAMM). The model parameters were fitted using Markov chain Monte Carlo simulations.

The model objectives were to assess the right size of the H5N1 epidemic, assess the impact of current mitigation measures on future outbreaks, identify critical epidemiological data needed to prepare for future outbreaks, and predict future outbreak hotspots.

Study results

Susceptible Infected Infected Infection Dynamics (SEIR) models (20,000 stochastic simulations) revealed that most current H5N1 infections in dairy cattle are concentrated on the west coast of the country. While the model was observed to overestimate case densities in some predicted outbreaks (Texas, Ohio and New Mexico), the model successfully simulated outbreaks for states with frequent reporting such as California, although it overestimated outbreaks in some other states (Texas, Ohio and New Mexico) as potential outbreaks in which researchers are in those states.

Alarmingly, only 16 of the 26 states where the model indicates a majority of simulations would have recorded an H5N1 outbreak by December 2, 2024 had actually reported one, indicating a high level of underreporting. Arizona and Wisconsin are expected to become future hotspots of H5N1 outbreaks. Indiana and Florida also have a significant risk of H5N1 outbreaks.

Research into current mitigation measures shows that they are not sufficient to control the prevalence of H5N1 in the country. Notably, the only current mitigation measure enforced in the states is exported cattle (screening up to 30 cows/herd for H5N1). Model predictions showed that increasing this screening to even 100 cows/herd would result in only a slight reduction in mean outbreaks and would not fundamentally change the trajectory of the epidemic.

Notably, the SEIR infection model does not account for other zoonotic viral reservoirs in model predictions. The ongoing epidemic for influenza and the possibility of these birds infecting livestock may worsen model predictions.

Conclusions

The present study and the SEIR model it presents suggest that current reports of H5N1 dairy cattle prevalence are an underrepresentation of the true concentration of the disease in the United States. Current anti-H5N1 transmission interventions are not sufficient to prevent additional outbreaks throughout 2025. With the highest risk of future outbreaks, Arizona, Wisconsin, Florida and Indiana require additional surveillance efforts.

“Significant increases in testing are urgently needed to reduce the uncertainty of model projections and provide decision-makers with a more accurate picture of the true scale of the national epidemic.”

Sources: