How ProPublica Investigated a Bird Flu Outbreak in America’s Heartland

Earlier this year, as I began to report on bird flu, an outbreak was raging across farms in Ohio and Indiana. Over 18 million hens - 5% of America's egg-laying chickens - were ultimately culled in a cascade of cases that, to me, appeared connected.

My reporting confirmed my hunch that they were linked: After the first farm was infected, the virus spread from that site to others, moving from farm to farm in a way that revealed a major flaw in America's strategy to combat bird flu.

On Tuesday, ProPublica published a story that demonstrates how the federal government is failing to control the spread of a deadly virus that could spark a pandemic.

The U.S. Department of Agriculture typically attributes bird flu outbreaks to failures of biosecurity - meaning farmers have not done enough to protect flocks from contamination by wild birds.

But my genomic analysis shows wild birds had little to do with this particular cluster of infections. Although the USDA said it tested nearly 1,000 virus samples in wild animals from December to April in Ohio and Indiana, no nearby wild birds were found infected with this outbreak's strain.

I did find a strong predictor of infection during the first few weeks of this outbreak: whether a farm was downwind from that first contaminated facility. That pattern reinforced the suspicions of egg producers and some local officials that the virus may be spreading on the wind.

If bird flu is airborne, the government's current biosecurity-based strategy cannot protect farms on its own. A poultry vaccine likely would have stemmed the damage from this outbreak, experts told me. Yet while other countries have curbed infections through vaccination, the U.S. has not authorized those efforts amid political and economic pushback.

The USDA told me it didn't investigate whether the wind contributed to the outbreak's spread.

Here's how I used genetic markers, satellite imagery, property records, trade notices, wind simulations and Google Street View to do the work USDA did not.

Pinpointing the Outbreaks

The USDA does not release identifying information about farms infected with bird flu - only the counties in which outbreaks occur. Between December and March, cases in Ohio and Indiana spanned six adjacent counties across more than 2,500 square miles in one of the nation's densest poultry-producing regions.

I worked out a way to pinpoint which farms had been infected by taking a close look at the notices the USDA publishes detailing regions restricted from exporting poultry products due to bird flu outbreaks. These notices describe a 10-kilometer radius around an outbreak to avoid revealing the actual site. They do give the northernmost point of the circle, and I found that plotting 5 kilometers south of that point revealed the site of the outbreak.

Knowing the actual locations led me to realize that the majority of cases clustered around the intersection of three counties, within a few miles of one another. It appeared much more likely that the virus was spreading among farms than repeatedly spilling over from wild birds.

Tracing the Virus

To find out if that was the case, I turned to genomic data from virus samples collected on farms and sequenced at the USDA's National Veterinary Services Laboratories.

The labs publish many of these sequences on GenBank, the National Institutes of Health repository for genomic data. Computational biologists and virologists visualize that data using a tool called Nextstrain, which builds phylogenetic trees showing how virus strains are related.

One way flu viruses like this one mutate is by swapping genetic code between strains in a process called reassortment. Each virus contains eight distinct segments that can mix and match when multiple strains infect the same cell, resulting in a new strain of the virus.

Researchers developed versions of the Nextstrain tool to visualize two specific genotypes of bird flu, the most common ones seen last year: B3.13, which was spreading in dairy cattle, and D1.1, prevalent in wild bird populations.

I altered the code of that software to target a less common strain of the virus, labeled D1.3 by the microbiologists at USDA's labs. There were two main lineages of D1.3 in the United States. One spread across hundreds of miles, killing off wild birds and poultry from Wyoming to Puerto Rico. The other lineage was made up of samples of the virus found only in domestic poultry and one human - exclusively in Ohio and Indiana.

I traced that family tree of viruses back to the root: a pair of samples that I was able to link to a farm called Howe's Hens, the first involved in this outbreak, in late December, just weeks before cases exploded in the region. (Howe's Hens" is how the farm is referred to in documents from the National Poultry Improvement Plan, which certifies egg exporters, but the farm is also referred to as Spring Meadows" by Cooper Farms.)

The samples from Howe's Hens had fewer mutations than the viruses that were sampled later on, and those later samples shared the unique mutations that first appeared at Howe's Hens.

One of the two samples bore a mutation that continued to show up in the other cases, while the other did not. The simplest explanation, virologists told me, is that as the virus spread through the farm, the mutation arose and persisted in other chickens infected on the farm. Because that mutation likely occurred on the farm, the farms that were subsequently infected were likely exposed to the virus that originated at Howe's Hens.

Virologists said that it was important to compare the virus in poultry to that found in wild birds, in case the virus was actually spreading in wild birds. But there were no viruses sampled from wild birds that bore any resemblance to this lineage. That's not to say that the infection didn't originally jump from wild birds into Howe's Hens. Rather, the government examines thousands of samples of the virus from animals - wild and domestic, dead and alive - to help researchers better understand how the virus is spreading and changing. But in all of those tests, they never happened across a source or close relative of the virus that caused this outbreak.

That was particularly surprising because Ohio and federal officials were pointing to wild birds as the main culprits spreading the disease, particularly a flock of wild geese found dead in a nearby wetland shortly before the outbreak started. However, documents I requested show that the geese were afflicted with an entirely different strain of the virus.

The question of how to control bird flu revolved around understanding how the flocks were infected in the first place. USDA, saying that wild birds were behind the vast majority of cases since it emerged in 2022, argued for stronger biosecurity practices to prevent contact between hens and infected birds, and to keep workers contaminated with the virus from infecting their flocks.

Farmers and industry leaders pointed to evidence suggesting a different source for the infections, saying that sick birds first appeared near air inlets and the virus infiltrated farms with seemingly impenetrable protocols to guard against outside contamination - arguing that potential airborne spread of the virus needed to be taken more seriously. I decided to look into it.

Blowing in the Wind?

Experts cautioned me to consider egg producers' motivations as I examined their claims of airborne spread: The industry was under intense pressure to solve the bird flu crisis by raising farms' levels of biosecurity to ensure that no wild birds or errant workers were bringing the virus into barns. Those higher levels of biosecurity are expensive and time-consuming, but wouldn't be very effective if the virus was spreading primarily through the air. Concluding that airborne transmission drove cases would relieve that pressure on individual operators and put more of a burden on the government for a fix.

The progression of cases certainly seemed like they could be spreading over the air. To evaluate the possibility, I tried to plot weather data collected from nearby airports, looking for patterns between cases and wind direction and speed. But airport wind records pointed in different ways at different speeds, and it wasn't clear what the pattern was.

So I turned to previous work on the airborne transmission of bird flu. Yang Zhao, an engineer and animal scientist at the University of Tennessee Institute of Agriculture, had researched other outbreaks, simulating where wind would have blown dust by using a model developed by the National Oceanic and Atmospheric Administration. That model, dubbed HYSPLIT (short for Hybrid Single Particle Lagrangian Integrated Trajectory) takes historical weather data as an input and simulates where particles would have traveled.

Mirroring those techniques, I generated a plume" that showed where dust originating from the first infected farm would have blown over a certain time period. I looked at the period when I was confident that there were infected birds on that farm: Dec. 22, when dying chickens were first noticed, to Dec. 31, when the entire farm had been depopulated.

This analysis doesn't account for how much infectious material would have actually been spreading in the environment. That would require additional information, like how many chickens were sick and at what rate. The USDA returned a public records request that likely would have contained some of those facts, but every piece of information was redacted.

The USDA responded to ProPublica's public records request with heavily redacted documents, including this HPAI Monitoring Depopulation Report, which detailed the steps taken to kill the birds on a farm once infection was discovered. Obtained by ProPublica via USDA

Instead, the plume represents the relative exposure: how much more exposure to dust a farm downwind would have had compared with another farm. It's also possible that it shows how wind would have influenced more than just the very small dust particles I simulated, like chicken feathers or insects.

Because we do not have reliable information on the order in which infections occurred, I could only investigate the possible spread from the first infected farm. Once the virus wound up on many farms, it became much harder to infer the path the virus took between them. In addition, the USDA redacted metadata from the samples taken on many of those farms that would have allowed some matching between farms and samples.

When I examined the plume that originated at Howe's Hens, there was a pretty clear overlap with the first wave of cases and where the wind was blowing over this period. But what if all of the farms in the area that were infected just happened to be in the direction that the wind was blowing? I wanted to get a better sense of whether this could be a coincidence.

To untangle factors that might be leading me astray, I built a list of all poultry farms in the surrounding area. I limited the analysis to a 12.4-mile radius around Howe's Hens (or 20 kilometers, the official surveillance zone around an infected farm where officials sample other farms to see where else an infection had spread).

There's no comprehensive database of poultry farms in the region, so I analyzed a database of building footprints created by Microsoft to identify groups of buildings that resembled the long, skinny barns typical in poultry production.

From the vantage of a satellite, hog farms can look similar to egg farms. So I used county property records that detailed the structures on parcels where I saw barns to sort out whether an image was of a turkey or chicken farm, versus one containing other livestock. This involved a lot of manual checking to make sure I didn't exclude any potential farms in the area, which required evaluating satellite and aerial imagery. In many cases, I also looked up addresses on Google Street View - most farms were right up against the road and often had some sort of sign showing what was inside.

In the end, I identified more than 200 poultry facilities in the surveillance zone, of which 55 were infected over the course of this outbreak.

I then overlaid the plume on those farms to get the relative level of exposure from the dust originating at Howe's Hens, to evaluate the relationship between that exposure and farm infections.

I found that farms with high dust exposure were about 20 times more likely to test positive for bird flu within three weeks. About 16% of the farms exposed to the plume were positive by Jan. 20, versus only one farm testing positive by then (0.8%) outside the plume.

However, as I drove around the region in May, it was hard to miss signs indicating egg and turkey operations were affiliated with Cooper Farms, a regional livestock company. Turkey producers in the area overwhelmingly contracted with the company, which produces meat largely destined for the deli aisle (versus whole turkey common on Thanksgiving tables), along with pork and eggs. According to the company, it also operates a massive feed operation, producing over a million tons annually, which supplies its many contract farmers. (Cooper Farms declined to speak to me and instructed all of its contracted farmers to do the same.)

Howe's Hens was affiliated with Cooper Farms. I expected there would be some common links between it and other affiliated facilities. After all, the USDA told me that it had attributed the spread of the virus in Ohio and Indiana to numerous movements of people and equipment."

Zhao, one of the airborne transmission researchers, said the data would be even more compelling if I could rule out Cooper Farms being the common link here. Using Street View and records from the National Poultry Improvement Project, I identified the farms exposed to the plume that were associated with the company.

Those results showed that the share of Cooper Farms infected in the first wave was roughly the same as the rest of the area's farms.

Zhao said that the results at least implied that something other than traffic between the farms was fueling the spread of the virus.

It's a good indication that factors such as airborne transmission may play a role, though more replication would be needed to reach firm conclusions," Zhao said.

What the Evidence Suggests

Experts were pretty aligned in saying none of what I had found definitively proves that the wind caused the virus to spread; there's a difference between correlation and causation, and the virus never follows just one mode of transmission.

It's hard, it's really hard - having spent my career studying viruses that spread through the air - to prove that the virus is in the air," said Seema Lakdawala, who studies the epidemiology of influenza viruses at Emory University. Yes, the wind might be important, but it may be other things that are moving in the wind, maybe the bugs, some other sort of vector moving along the wind that we have not accounted for."

Nevertheless, other experts I discussed our analysis with - eight in total - agreed that this data presented a plausible case for the wind playing a role in this outbreak.

It just seems so likely to me that this was an airborne thing," said Brian McCluskey, former chief epidemiologist with USDA's agency that oversees the response to bird flu. I mean, how else would it have moved around so quickly?"

Zhao said that ProPublica definitely moved one step further from our research in the past."

This result is very exciting, very interesting to me," he said.

Read More

What the U.S. Government Is Dismissing That Could Seed a Bird Flu Pandemic

The USDA insisted that this particular outbreak was unique" and not representative" of the entire wave of bird flu that started in 2022, and that the overwhelming majority" of infections stem from wild birds. The agency said its biosecurity strategy remains rooted in real-time data, internationally recognized best practices and a commitment to transparency and continuous improvement," and that it is proactively assessing" the possibility of vaccinating poultry for bird flu.

Experts told me that understanding what drove this massive outbreak was important, and it didn't seem like USDA was doing that work. The agency did not evaluate airborne transmission in this outbreak. It also doesn't make it easy for others to do so, withholding key information that would allow journalists and researchers to evaluate the spread of the virus.

As infections surge again this fall, the USDA continues to urge farmers to improve biosecurity while it dismisses a significant way the virus could be spreading.

The post How ProPublica Investigated a Bird Flu Outbreak in America's Heartland appeared first on ProPublica.