Avian flu outbreak risks in 2023

Fears of a bird flu pandemic have been growing. Some influenza viruses are endemic in humans, but while those strains of the virus come and go, causing significant waves of illness and death in flu seasons, they don’t break out and become full-blown pandemics. The risk of a new influenza pandemic primarily arises from the possibility of a strain spreading to humans from other animals, and adapting to effectively spread between us.

That doesn’t happen easily: usually, animal viruses are poorly suited to human respiratory systems and don’t infect us, or if they do they don’t start spreading between humans. But at the moment, the world is seeing its largest-ever outbreak of H5N1, and has seen large outbreaks of other avian influenzas (such as H5N8) in recent years. That means more possible virus transmission between birds and humans, or between birds and other mammals (and then between those mammals and humans). And that means more opportunities for those viruses to mutate and become human-transmissible.

A new pandemic would be devastating, especially since bird flus are often much deadlier to humans than Covid, and more likely to kill people of working age and thus do more economic damage. Different strains of avian influenza have infected humans in the past (and there is evidence of limited spread between family members) but so far there has not been any sustained human-to-human transmission, as far as the world knows.

We wanted to look at how likely it was that the virus would adapt to effectively spread between humans, and, if it did, what the likely impacts would be – in terms of continued spread, likely deaths, and policy responses.

Direct impacts

Will there be at least one confirmed case of human-to-human transmission of any avian influenza before August 2023?

The first thing to do is establish a base rate. What’s the probability of seeing human-to-human transmission of bird flu in a randomly selected six-month period? Forecaster 1’s method was to look at the World Health Organisation’s timeline of H5N1 events (H5N1 is a subtype of avian flu, and the virus responsible for the current bird flu outbreak). It found six possible or probable incidences of human-to-human transmission, all between 2004 and 2008.

Taken at face value, that’s six events in five years, or 1.2 events per year. It’s about six months until August so that suggests the probability of seeing a possible incident in that time is about 60%.

But those are possible incidents. Forecaster 1 imagined that at least half of them would not be confirmed by health authorities in 2023, putting the base rate at a maximum of 30%. Then they looked for reasons to move that up or down. Reasons to move up included the large and ongoing avian flu outbreak, and the fact that it looks like there has been at least some mammal-to-mammal transmission, notably in a mink farm in Spain.

On the other hand, while those six incidents all took place in five years, the WHO was monitoring both before and after: “The time window should probably be from at least 2003 all the way to 2014, giving a base rate of less than 14%,” Forecaster 1 said. Also, avian flu is not usually as transmissible as human flu, and none of the staff at the mink farm seem to have caught it. Overall, Forecaster 1 gave a 15% chance of seeing one case of human-to-human transmission.

Other forecasters broadly agreed: Some went as low as 8% and some as high as 25%, but all thought it was relatively unlikely. Forecaster 2 gave a figure of 18%, saying “I'm going just slightly over what I'd consider to be a base rate, due to the current H5N1 outbreak and likely increased surveillance being more likely to positively identify a case”. Forecaster 7 says that the "bird-mammal interface is growing, so we should expect more mammals get infected" and believes the risk is reasonably high (at 21%) since previous research has shown how H5N1 is able to quickly gain adaptations when it infects mammals.

Will we see 100 human-to-human cases by August?

The forecasters were asked this question on its own, but also conditionally: How likely are we to see 100 human-to-human cases given that we see one such case?

The forecasters felt the second scenario was plausible, albeit not likely. Avian influenza might not be as infectious as Covid-19 or ordinary endemic flu, but that doesn’t rule out wider spread. “There were tens of thousands of cases during the Ebola outbreak of 2014-16, despite Ebola not being particularly transmissible and close contact being required,” Forecaster 3 said. On the flip side, none of the six “possible” cases of human-to-human transmission of H5N1 in the past have led to larger outbreaks like this, Forecaster 1 noted.

Those two forecasters also disagreed over whether public health systems and responses would likely be able to stem any outbreak. “Public health authorities would be likely to prevent more than 100 cases as there have been a number of probable human-to-human cases in the past,” said Forecaster 3, while Forecaster 1 said “I honestly do not think our disease surveillance and response is effective enough to stop an influenza virus from infecting 100 people once there's community transmission”.

The average estimate hides a wide spread on the conditional question: the lowest figure was 4.2%, while the highest was 41. The range in the no-condition question was much smaller, with no forecaster going above 7.5% and most clustering around 1%. That makes sense: the probability of an outbreak of more than 100 cases should be the same as the probability of one case multiplied by the probability of 100 cases given one case, and that is indeed what the numbers are.

Will we see more than 50,000 deaths by January 2024, if there is one case of human-to-human transmission before August 2023?

Again this is a conditional question: given that we see at least one case, what’s the probability of 50,000 deaths before the end of the year?

Forecaster 1 pointed out that 50,000, although it sounds impressive, isn’t actually a huge number, assuming the disease goes global. “It’s only two to five times the number of deaths in a regular UK flu season,” they said. But they felt it was still unlikely: “H5N1 seems to have been transmitted between humans at least a couple of times in the past couple of decades,” they said. “Seeing it again wouldn't necessarily freak me out”.

Forecaster 2 pointed out that if there was an outbreak, initial cases would be rapidly quarantined, and countermeasures such as social distancing deployed. Vaccines for H5N1 flu exist, although their efficacy is unclear and stockpiles might not be sufficient. “On the other hand,” they said, “the case-fatality rate (CFR) of H5N1 is very high – around 50%, much higher than that of SARS-CoV-2, even in the early stages of the Covid outbreak when many of those infected weren't being tested, which suggests that the infection fatality rate (IFR) is also going to be substantially higher than that of SARS-CoV-2.

“There are easily more than 50,000 deaths globally in a typical seasonal flu outbreak, and seasonal flu has an IFR of less than 0.1%”. Nonetheless they felt it was “more unlikely than likely” that one human-to-human case would lead to 50,000 deaths, and put a 16% figure on it, one of the higher estimates. Most forecasters put the figure at more like 5%, although one went as high as 30%.

If there isn’t a confirmed case of human-to-human transmission by August, the forecasters felt the probability of seeing 50,000 deaths by January was less than 1%.

Will we see more than 50,000 deaths by January 2024, if there are 100 cases of human-to-human transmission before August 2023?

(The same as the above, except for 100 cases instead of one)

If there are 100 confirmed cases of human-to-human transmission, the forecasters felt, that meant it was pretty likely there’s sustained community transmission. “My intuition is that this would be a strong signal that something very bad is happening,” said Forecaster 4. “It would demonstrate that either the first line of defence in animal monitoring and prevention has failed, or the virus has emerged elsewhere and transmitted rapidly”.

There are scenarios where 100 cases doesn’t lead to wider spread: “If it broke out in a meat-packing factory or something, where the conditions happen to be very well-suited to spreading an airborne virus,” said Forecaster 1, it might spread rapidly there but be unable to take hold in society at large. They also pointed out that SARS-1 reached thousands of cases but “petered out quite quickly”. Nonetheless, “Intuitively, I feel pretty sure 50,000 deaths is more likely than not under this scenario”.

The forecasters had a wide range, from around 20% to around 70%, representing considerable uncertainty. “We've not seen 100 human-human cases of avian flu before,” said Forecaster 1, “so this is mostly down to intuition for me”.

If there are not 100 human–human cases before August, it is still possible for there to be 50,000 avian flu deaths before the end of the year. As Forecaster 3 puts it: "there is a slight chance that there hasn't been enough time to reach 100 cases before August (but that it is reached after August and, due to exponential growth, there are quickly hundreds of thousands of infections)". Additionally, as Forecaster 1 argues, it could spread solely as "a vector-borne disease, with regular transmission from birds (or other animals) to humans, but this strikes me as very unlikely (~1/1000)".

Policy responses

Will the US Federal Reserve’s interest rate be above 2% before 15 October 2023, if there are 100 cases of human-to-human transmission before August 2023?

The Federal Reserve has been hiking interest rates to combat inflation for the last few months. But the threat of a new pandemic might cause turmoil in financial markets and a liquidity crisis: in March 2020 the Fed cut interest rates by 1.75%. We wanted to ask whether something similar would happen again, were bird flu to start spreading among humans.

Some forecasters strongly expect rates to be cut in the face of a serious bird flu outbreak. “If we did see more than 100 human-human transmission cases,” said Forecaster 5, “it would probably be part of a huge pandemic. That would cause substantial economic turbulence, and rapid cuts to interest rates as central banks sought to avert financial market panic”.

Forecaster 4 agreed. “More than 100 transmissions is likely to be in the ‘this is very bad’ bucket, so there is time for the follow-on economic shocks to kick in for the Fed to move below 2%”.

On the other hand, “World markets reacted quite late to the Covid outbreak, although I’d be surprised if that happened again,” said Forecaster 1, “and there’s a chance it looks like SARS-1 or takes more than four months to end up looking really bad”. Also, said Forecaster 2, the Fed’s recent rate rises have given it room to respond without going below 2%: the federal funds rate is currently about 4.5%.

On balance the team felt that even given 100 cases, there was enough room for the Fed to manoeuvre, and enough scenarios in which 100 cases didn’t necessarily mean another Covid-style disaster, that the rate would likely stay above 2%. The range was considerable, though, from near-certainty to below 30%.

If there aren’t 100 cases, the forecasters felt, the rate is very unlikely to drop below 2%, unless there is some other shock to the economy, such as a nuclear detonation in Ukraine.

Will at least three countries impose border quarantine policies to prevent the entry of avian influenza by 1 January 2024, if there are 100 cases of human-to-human transmission by August 2023?

Shutting borders was controversial in some countries during the early weeks of the Covid pandemic, but in the end most did it. Given the threat of a new pandemic, would they do it again?

Some would, the forecasters felt. “I think you can get to three countries quite quickly here,” Forecaster 4 said. “Taiwan, Japan, and Singapore for sure due to their governments and relative success with doing so before. If this goes above 10 or 20, it’s more interesting”.

In a post-Covid world, also, countries will want to make it clear that they’re doing something, Forecaster 6 said: “I expect border quarantine policy to a) be relatively cheap to implement politically b) appeal as a way to signify taking COVID style risk seriously”.

Forecaster 7 also felt countries would have learnt from Covid. “I think from Covid it's clear that in a pandemic border closures are very effective, and at a minimum they buy time to get vaccines, develop therapeutics, etc,” they said. “I think some countries would be quick to respond to the early signs of what could be a pandemic, and take the correct decision to slow or prevent the spread to their countries”.

All the forecasters felt it more likely than not that at least three countries would close their borders if there were at least 100 detected human-to-human cases by August, with the lowest estimate 62% and the highest 97%.

If there were fewer cases, they thought it was much less likely, though not vanishingly so: “I could still see multiple countries being extra-careful even before there was large amounts of spread,” said Forecaster 8, “perhaps as soon as there were a few confirmed cases of human-to-human transmission”. Forecaster 2 noted that they could be closed to protect the poultry industry. Still, the highest forecast was 22% and the median was 7.2%.

Will the US start an expedited vaccine process for avian flu by January 1, 2024, if there are 100 cases of human-to-human transmission by August?

The sped-up process for researching and licencing vaccines for Covid undoubtedly saved many lives, and started before most people even realised Covid was a serious threat. Will the same happen if it starts to look as though bird flu is on the rise?

Probably, the forecasters felt, not least because there’s a US presidential election on the way. “This would be such an easy political win, and I expect there would be a lot of lobbying for it if things start looking more scary here,” said Forecaster 1. “The Democrats seemed to think that it was very important that the Covid vaccines were announced after the US Presidential Election, because voters do respond positively to these kinds of things, and I can't imagine they would ignore a straightforward policy win when the opportunity arises”.

Forecaster 3 agreed: “The Biden administration will want to be taking every precaution. The election could hinge on their response to a potential outbreak of avian influenza”.

The counterargument, some forecasters noted, was that some vaccines exist already. But all but one of them need to be incubated in hens’ eggs, “which is a bit inconvenient when it's an avian flu,” said Forecaster 1. “I strongly expect mRNA vaccines would be designed and approved rapidly”.

There was a dissenting voice: Forecaster 8 was very pessimistic on the governmental response. “Unfortunately, the odds of this are quite low,” they said. “The US government have been hesitant to engage in expedited vaccine processes for next-generation COVID vaccines that would protect better against the current strains of COVID, so I think America would only do this if it looked like we were going to see a pandemic that was expected to kill at least hundreds of thousands of Americans”.

For an introduction to forecasting — and to improve your ability to reason about different scenarios — sign up to do a workshop with us Swift Centre Workshops.