How to pinpoint the H5N1 mortality rate in humans

Avian flu is reported to kill more than half the people it infects. The reality is much more complicated

Published July 23, 2024 12:36PM (EDT)

Farmer moves her cows into a barn for their evening milking on April 25, 2017 near Cambridge, Wisconsin. (Scott Olson/Getty Images)
Farmer moves her cows into a barn for their evening milking on April 25, 2017 near Cambridge, Wisconsin. (Scott Olson/Getty Images)

This article originally appeared on Undark.

Just how deadly is the H5N1 avian flu? The virus, which is currently sweeping through U.S. dairy herds, rarely jumps to human beings, at least for now. But when it does the consequences can be grave: The World Health Organization reports that 52 percent of people known to be infected with H5N1 have died from the disease.

The figure has been widely cited in academic papers, public health communications, and media reports, where it can provoke apocalyptic visions. “Bird flu pandemic could be ‘100 times worse’ than COVID,” claimed one New York Post headline. An article in The Guardian leads with the WHO’s “enormous concern” about the spread of H5N1, which, according to one lead scientist quoted, has an “‘extraordinarily high” human mortality rate.

The actual picture, while still alarming, is more complicated. The WHO’s H5N1 mortality figure, an average of wildly different death rates from past outbreaks, doesn’t factor in mild cases that went undetected. Even less certain is how lethal H5N1 would be if it evolves to spread not just from animals to humans, but also from person to person.

That genetic twist would likely diminish H5N1’s virulence, experts predict, but no one can say how much less deadly it might become. And even a virus that kills far fewer than 52 percent of people would be devastating: As the world saw with the Covid-19 pandemic, even a death rate of 1 to 2 percent can be catastrophic.

But answering how lethal an H5N1 pandemic might be is no easy feat. A dive into that question reveals the ongoing challenges — and, some experts say, failures — of tracking the virus. And it offers a glimpse at the difficulty of communicating the risks and unknowns about an emerging pathogen.


Since the first human outbreak of H5N1 in Hong Kong in 1997, the disease has cropped up sporadically around the world, almost entirely infecting people who worked directly with poultry. Between Jan. 1, 2003 and May 3, 2024, the World Health Organization recorded 889 cases of H5N1 and 463 deaths. Dividing the total deaths by the number of cases results in what epidemiologists call a case fatality rate, or CFR, of 52 percent.

But CFRs are notoriously uncertain. “The fundamental problem is that a case is not a tightly defined scientific concept,” said Harvard epidemiologist Marc Lipsitch. “These numbers are something we have to use because we don't have something better, but people in the business are aware that they are potentially deceptive.”

When and where researchers look for cases can heavily bias CFRs. A virus that produces mild, undetected infections in 998 people, sends two people to the hospital, and then kills one of the hospitalized patients will have a CFR of 50 percent if public health authorities only manage to detect those two serious cases. But the true fatality rate would be one person in 1,000, or 0.1 percent.

Accurate CFRs are critically important in an outbreak because marshalling a public health response depends on understanding the disease’s severity. For example, when H1N1, also known as swine flu, emerged in Mexico in the spring of 2009, tens of thousands of mild cases went undetected, causing health authorities to overestimate the severity of the disease. In a study published later that year, Lipsitch and an international group of researchers from organizations including the Centers for Disease Control and Prevention estimated that the actual case count among Mexican residents that spring was about 100 times higher than officially reported.

On the flip side, though, if researchers overlook fatal cases they will underestimate the lethality of a virus. For instance, research suggests that health authorities initially undercounted deaths in a 2003 SARS outbreak in Hong Kong because they didn’t follow patients long enough to record everyone who died of the disease.

Like many experts, Peter Palese, a microbiologist at the Icahn School of Medicine at Mount Sinai in New York, thinks that the CFR of 52 percent for H5N1 calculated by the WHO likely overestimates the disease’s severity. To meet the WHO’s definition of an H5N1 case, the person must have had a fever and tested positive for the virus in a lab with the technical capacity to follow WHO protocols. Because many of the outbreaks have been in rural areas without sufficient testing capabilities, the case count is drawn almost exclusively from patients who were sick enough to be hospitalized. Meanwhile, said Palese, many milder infections likely went undetected, although the exact number of those silent infections is unknown.

“It is not completely clear whether these high fatality rates are real,” said Palese.

“These numbers are something we have to use because we don't have something better, but people in the business are aware that they are potentially deceptive.”

Outbreaks are like an iceberg where serious infections are immediately visible, but the larger numbers of mild infections are out of sight below the water line, said Malik Peiris, a virologist at the University of Hong Kong with extensive experience with H5N1. One of the best ways to get a more accurate case count, he said, is to test community members’ blood for antibodies against H5N1, which would indicate a previous infection: “That gives you a much more accurate picture of the bottom of the iceberg.”

Researchers have conducted dozens of such studies. But results from that research, said Peiris, “are a bit mixed and somewhat confusing.” While antibody studies of some H5N1 outbreaks find evidence of widespread mild infections, studies of other H5N1 outbreaks do not, even among people who worked closely with infected birds. Peiris described the disparity as “rather puzzling.”

It's possible, said Peiris, that antibody tests miss some cases. Type A influenza viruses such as H5N1 are characterized by the combination of two proteins on their surface: hemagglutinin, which can be one of 18 types numbered H1 to H18, and neuraminidase, numbered N1 to N11. Compared to the H1 and H3 proteins in the influenza A viruses responsible for seasonal flu, H5 proteins trigger a weaker response from the immune system, said Peiris: “People may be getting mildly infected, but it’s not enough to elicit an antibody response.”

Peiris and other experts described the current H5N1 outbreak in dairy farms as a prime opportunity to investigate how and where H5N1 is spreading as well as how we might contain it. But in many areas, farmers who are worried about the threat to their livelihood won’t allow officials on site to test workers or animals.

As of July 18, H5N1 has been identified in 163 herds of dairy cattle in 13 states. But wastewater surveillance data showing spikes of Influenza A outside of flu season in some regions suggests that H5N1 could be circulating more widely, said epidemiologist Michael Osterholm, who heads the University of Minnesota’s Center for Infectious Disease Research and Policy. “We’ve already missed a big chunk of potential worker infections,” he said. Still, even now, antibody testing would give us a "darn good picture” of the number of human cases.

“That’s the kind of thing we really need to get a handle on,” Osterholm said. “The absence of evidence is not evidence of absence.”

Between March 2024 and now, H5N1 surveillance has entailed monitoring about 1,570 people who have been exposed to infected animals and testing at least 62 people, CDC spokesperson Jasmine Reed wrote in an email to Undark.

“It is not completely clear whether these high fatality rates are real.”

There have only been 11 reported cases of bird flu in humans in the U.S. since 2022, according to the CDC, with just five of those confirmed as H5N1. All cases have involved farmworkers who worked with infected animals. The agency is also providing technical assistance on an antibody testing by the Michigan Department of Health and Human Services that is looking for asymptomatic infections among people who worked with sick cows.

But, like many experts, Osterholm is worried that testing is wildly insufficient. More than 9 million cows produce milk across all 50 states, according to the U.S. Department of Agriculture. And the dairy farm industry employs more than 100,000 people.

While Michigan appears to be ground zero for the current outbreak, Osterholm said, “I don’t believe for a second that’s really true.” Michigan’s agriculture and health departments have just been more proactive about surveillance and testing, he said: “I'm convinced, quite honestly, that if they can get on more farms in more states, you'll see this is much more widespread.”


Accurately assessing the lethality and spread of H5N1 is crucial to predicting what could happen next.

For now, H5N1 has proven deadly, but still hard for humans to catch. “If this were to become a pandemic virus,” Osterholm said, “it would have to go through major changes.”

What those changes would mean for the virus’ lethality, though, is unclear.

H5N1 could develop the capability for person-to-person transmission in a few ways. In the process of replicating, viruses could acquire random mutations that make them better suited to a human host. In addition, different types of viruses can swap genes through a process called reassortment. So, if a human or other animal were infected with both a typical human flu virus and H5N1, those viruses could generate a new strain that was both deadly and easily transmitted.

Only a small set of avian influenza viruses have evolved to infect mammals, said Thomas Friedrich, a University of Wisconsin virologist who studies the evolution of pandemic viruses. To infect a host, viruses latch on to receptors on the surface of cells, Friedrich explained. The receptors H5N1 binds to in birds are configured differently from most of those in humans. People only have bird-type receptors deep in the lungs, said Friedrich, where infection is associated with severe disease.

“That can help explain both why human infections with H5N1 viruses have tended to be very severe,” he said. “And why those viruses that infect those unfortunate humans have a hard time getting from that human to another one.” To efficiently spread from person to person, the virus would need the ability to attach to human-type receptors in the upper respiratory tract. Once it takes hold there, talking, sneezing, coughing, and even breathing will then spew it into the world.

While it would probably only take a couple of genetic changes to get to that point, said Friedrich, “we don’t find a whole lot of evidence that bird viruses infecting humans are evolving toward the ability to bind those upper respiratory tract cells.”

One theory for why, so far, H5N1 has not evolved to infect the upper respiratory tract in people is that the virus so successfully survives and replicates in the lower lungs that it outcompetes any mutations, said Friedrich. Data from his lab and others suggest that mutations that could bind with human-type receptors die off before taking hold.

“I'm convinced, quite honestly, that if they can get on more farms in more states, you'll see this is much more widespread.”

But that could change, he said, when the virus infects a species with both human-style and bird-style receptors. For example, researchers have pinpointed the start of the 2009 H1N1 pandemic to pigs, which can be infected with both human and bird flu. And a recent study in preprint that has not yet been peer-reviewed finds that cow udders can also contain both types of receptors and so could potentially become a mixing vessel for bird and flu viruses.

A new study published in the journal Nature suggests that may already be happening. A team of researchers from the U.S. and Japan found that H5N1 virus isolated from the milk of infected cows could bind with both human and bird receptors. Those results are controversial, however, as other researchers who’ve studied current strains of the virus concluded that it hasn’t become more specific to humans.

In the last two years, H5N1 has spread to nonhuman mammals such as foxes, skunks, cats, mice, and marine mammals — perhaps both because they are encountering more infected birds and because the virus has become better suited to mammalian hosts, said Friedrich. If the virus further evolved to infect the upper respiratory tract, rather than the lower lungs, of humans, researchers speculate that could make it less lethal, he said: “But there is no hard-and-fast rule that viruses don’t evolve to kill their host.”

Like many researchers, Peiris is concerned that if H5N1 becomes a pandemic virus, the mortality rate would be much higher than that of Covid-19. He pointed to a recent CDC study showing that an H5N1 virus isolated from a person infected in a recent outbreak was lethal to ferrets, which he said are the best animal model for human severity and transmission.

An H5N1 pandemic “would have catastrophic consequences,” he said. “I have no doubt about that.”

There is one hopeful note. In early 2024, Peiris and his colleagues published a study suggesting that previous infection with the H1N1 swine flu may provide some protection against H5N1. In testing blood samples collected from a random sample of 63 adult blood donors, the researchers found that antibodies resulting from a previous infection of swine flu also reacted to the N1 protein in H5N1. While that immune response wouldn’t block an infection entirely, it might mitigate its severity, said Peiris. The team is now studying that possibility in animal models.


For now, many public health experts remain frustrated by the lack of clear data on H5N1 — especially following similar problems in the early days of the Covid-19 pandemic.

Public health authorities should be doing far more testing for evidence of H5N1 in agricultural workers, said Jennifer Nuzzo, who directs the Pandemic Center at the Brown University School of Public Health. But even if they did test more, said Nuzzo, there is a need for standardized protocols.

“This is one of those things that we know we need to do,” said Nuzzo, who spoke with Undark in June. Since then, the CDC has published such protocols for antibody testing, which would make studies easier to compare because researchers have used different methods in previous outbreaks. That has been an issue in the past, when, Nuzzo said, “we jumped to very consequential conclusions based on these data that could very well be the product of a very biased study design.”

While Nuzzo would like to see more farmworkers tested regularly, she acknowledges that it’s a hard-to-reach population. Farm owners aren’t always cooperative. And the workers, many of whom are undocumented, may also be reluctant to submit to testing that they view as a threat to their tenuous lives in the U.S. In the meantime, Nuzzo is adamant that farmworkers should be offered vaccination against the virus.

Amid the uncertainty, some public health experts suggest, the public conversation about H5N1 has become disconcertingly contradictory, with reassuring messages that risks are low juxtaposed against warnings of a brewing pandemic.

Communication about the threat of H5N1 often lacks nuance and perspective, said Osterholm. Figures like a 52 percent death rate, he said, do little to capture the profound unknowns about an ever-changing virus. At the same time, statements saying that there’s little reason for the public to worry about H5N1 — like recent pronouncements from the CDC — appear to downplay the threat. For example, Nuzzo emphasized that the risk to farmworkers is not low.

It’s true that the virus currently poses little risk to the general public, said Osterholm. “But all that could change tonight.”

Amid the uncertainty, some public health experts suggest, the public conversation about H5N1 has become disconcertingly contradictory.

Many health authorities view the flood of alarming and conflicting information on Covid-19 as an example of how not to communicate during a pandemic. Public guidance from the CDC was confusing and overwhelming, according to a 2022 internal review.

A common mistake was oversimplifying information, stripping out essential details and glossing over unknowns, said Nuzzo. That undermined people’s trust in advice that changed along with the evolving scientific information. “You have to take people on the journey with you,” she said. “Because if you put a fairly high-consequence conclusion in front of them and don't kind of have anything to back it up, I think it's natural that people are going to feel skepticism.”

The public is much smarter than they’re given credit for, said Nuzzo, “And I think people are hungrier for more information, not less.”

This article was originally published on Undark. Read the original article.


By Teresa Carr

MORE FROM Teresa Carr


Related Topics ------------------------------------------

Bird Flu H5n1 Health Pandemics Public Health Undark