Hide table of contents

Hi everyone,

I've been reading up on H5N1 this weekend, and I'm pretty concerned. Right now my estimate hunch is that there is a 5% non-zero chance that it will cost more than 10,000 people their lives.

To be clear, I think it is unlikely that H5N1 will become a pandemic anywhere close to the size of covid.

Nevertheless, I think our community should be actively following the news and start thinking about ways to be helpful if the probability increases. I am creating this thread as a place where people can discuss and share information about H5N1. We have a lot of pandemic experts in this community, do chime in!





Group of H5N1 manifold markets: https://manifold.markets/group/h5n1-bird-flu 


Plan for action

Fight status quo bias

In January 2020, many in the effective altruism and rationalist communities had correctly gauged the seriousness of the pandemic threat and were warning people publicly about it. Despite being convinced it was likely to become a pandemic I almost entirely failed to act beyond a few symbolic gestures such as stocking up on food/masks and warning relatives.

I consider this to have been the biggest personal failing of my life. I could have started initiatives to organize and prepare, I could have invested in mRNA producers, I could have researched how it would affect third-world hospitals. Yet all I did was sit idly by and doom scroll the internet for news about covid.

My goal with this thread is to avoid making that mistake ever again, even if it means most likely looking really stupid in a few months time.

How can we lower the chance of a serious pandemic?

I encourage everyone to think about actionable steps and be ambitious in their thinking. As far as I understand mink-to-human transmission is currently the primary reason to be concerned. What ways are there to minimize the chance of this occuring?

The following companies currently own vaccines for H5N1: 

Sanofi SAAflunov
GSK plcQ-Pan H5N1 influenza vaccine
CSL LimitedAudenz (and 1-3 more I think?)
Roche Holding AG Genussscheineoseltamivir (aka Tamiflu, not a vaccine), this one seems less useful than the others

Could we pay them to start scaling up production tomorrow? One thing to note is that all these vaccines are egg-based. Are mRNA vaccines possible to create for this? If so, what can we do to speed up the process of making them?

Any other ideas?

Sorted by Click to highlight new comments since:

For what it's worth, several of us at Alvea have been keeping a close eye on this for a while. If this escalates, then it is very likely that we would respond appropriately. (Note: My opinions are my own and I do not speak for Alvea in any official capacity.)

Can you speak on this since Alvea has been wound down?

As a layperson, I'm increasingly convinced that we're on a path to a devastating H5N1 pandemic within the next few years, possibly much sooner (writing this in late June 2024). My concern is based on the amount of "biocomputation" that is going towards / will soon go towards parallel search for mutations that have the side effect of triggering a pandemic in humans, plus the risk of viral reassortment with human-native viruses. I've been reading a lot about this over the past few days, and I'm eager to defend this position or revise my opinion in light of new arguments.

Here's my basic reasoning:

Clarifying my position here, I'm not claiming there will be a pandemic next week or next month. My claim is more that "common in cattle but not common in humans" is an "unstable evolutionary equilibrium" for H5N1, which is unlikely to persist for an extended period of time.

Another way of putting it is that the ambient risk level for H5N1, in the sense of the number of possible-pandemic dice which are getting rolled every day, seems to have gone up a lot recently. The risk level is likely to stay high, because I expect H5N1 to become endemic among cattle (if it isn't already). And regulators seem to be dropping the ball, just like they did for COVID. Calling your congressperson and telling them to ask regulators hard questions could be high-impact.

Thanks for tracking this, I definitely think being aware of + extremely up-to-date on "novel pandemics" is one of the things that our community (~GCR-focused EAs) should have at least a few people working on, though it's a hard thing to do without either under- or over- shooting. 

(It's possible these people already exist and are working quietly. In which case, great! I'm not sufficiently in touch with the space to know either way).

I'm not sure about keeping up to date on potential "novel pandemics" is a great use of resources for above reasons (see reply to Ebenezer). I would like to see us predict in real time even one potential virus moving from animal transmission through human to human pandemic first before committing huge resources.  I feel like "general pandemic prevention" stuff (stockpiling, UVA, better diagnostic tools etc.) as well as tracking virus's we already know can cause epidemics (like Ebola) is maybe a better use of resources but am very uncertain.

Its a super interesting discussion though.

Thanks I absolutely love this list and absolutely agree with your reasoning, that points towards H5N1 being one of the most likely (if not the most likely ever identified) situation where a known pathogen could move from a minor issue to disastrous pandemic.

In saying that I still think its very unlikely, based on prior evidence that we will see this happen in real time. I'm not sure that humans have ever actually followed a specific virus that then became a dangerous pandemic. Can you think of an example that fulfils these critreia?

1. Virus identified in advance that was either non-transferrable to humans, or (like H5N1) has very limited human transmission
2. Prediction made that the virus could become dangerous
3. The virus mutates and becomes dangerous, causing an epidemic/pandemic

Previous dangerous diseases that emerged from other animals (HIV, Ebola, Covid, Swineflu) were not predicted in advance.

Because of that I would rate this statement as quite an overstatement."we're on a path to a devastating H5N1 pandemic within the next few years, possibly much sooner". The most likely scenario is that we never get a H5N1 pandemic.

This doesn't mean we shouldn't be spending far more money on the issue and focusing on it, there's obviously a real chance that H5N1 becomes disastrous, I just think well below 50%. I in general though rate priors very heavily, far more heavily than theory so it depends on your prediction methods.

This is a pretty good summary here by the institute for progressalso, where they estimate risk at 4% in the next year, my instincts are it might be even lower. I like their cascade of probabilities, but at a few stages I would have gone with lower probabilities.

We could also ask serious forecasters here what they think? @Peter Wildeford @NunoSempere 

Thanks for the pushback, Nick.

No, I can't think of any examples that meet your criteria. As a layperson I wouldn't know about them if they existed, anyway.

I could quibble with your implicit method for computing a prior, however. You mention 4 zoonotic pandemics that were not predicted in advance. I'd argue the correct denominator here is pandemics which were predicted in advance. How many times in history have we had an argument this strong for a pandemic which was ultimately a nothingburger? That should be our reference class.

The joke goes that "economists have predicted 9 of the last 5 recessions". Have epidemiologists predicted 9 of the last 0 pandemics? Furthermore, how does the emergence of modern sequencing technology factor in?

Note that I am not an epidemiologist! So that outside view might not be worth much. I'm a computer scientist, and I'm thinking about this very computationally and probabilistically. See my terminology: "biocomputation", "parallel search for mutations", dice rolls, and perhaps an implicit reference to hill-climbing (H5N1 got knocked off its previous local maximum and now it's searching for a new one). I suspect many epidemiologists do not currently grok this computational/probabilistic argument, which is part of why I wrote my comment. But I know there are big parts of the epidemiological picture that I'm missing.

Anyway, here's another reference class that might be interesting: Endemic viruses in domesticated mammals. Are there cases where they persisted for many years without becoming endemic in humans? Perhaps canine parvovirus? It's unclear how to apply this outside view to a virus like H5N1 which (a) has managed to infect so many species, (b) is known to infect humans sometimes, and (c) is currently evolving to better infect mammals (may have lots of low-hanging fruit to capture).

If one wanted to make a serious effort to forecast this, one can imagine a Monte Carlo model which accounts for the H5N1 mutation rate, average case load in US cattle, total virions per infection, fitness advantages and disadvantages associated with various mutations, human contacts, etc. My basic intuition is that many plausible parameterizations for this model will produce a human pandemic some time within the next few years. Maybe there's some way of doing a quick and dirty monte carlo using Guesstimate.

Thanks yeah I agree with your first question being important, and I would say we have predicted Zero new problematic pathogens ever. I agree with you that the numerator could be pandemics from new zoonotic pathogens which were predicted in advance, and that number I think is zero, making it hard to calculate a prior with this....

You make a decent argument that we could look at pandemics from organisms which we already know have pandemic potential in humans and then see how many of those we got correct. I don't know the answer to this, but I would imagine its a VERY LOW number (or even zero) again.

My feeling is that there are so many animal diseases out there, the scenarios where diseases combine or mutate to form variants that are dangerous to humans is so hard to predict and random, that even the majority that appear "super close" to being dangerous will never actually become dangerous.

Predicting pandemics is not like predicting volcanic eruptions - at least not yet and you are right that sequencing and other technology will gradually make us better at this - I just think we aren't nearly there yet.

"If one wanted to make a serious effort to forecast this, one can imagine a Monte Carlo model..." This is basically what the institute for progress did, but in a more simple linear way have you had a look at their calculations?


Thanks for the reply.

My feeling is that there are so many animal diseases out there, the scenarios where diseases combine or mutate to form variants that are dangerous to humans is so hard to predict and random, that even the majority that appear "super close" to being dangerous will never actually become dangerous.

I suppose one could create a dataset of various viruses with columns like: Is it endemic in domesticated mammals? How many humans have been infected? Does it appear to be evolving? Can it reassort with human influenza? Etc. etc. and then train a regression or something to predict a pandemic. I suppose data would be extremely sparse, but if you rework the task as "predicting whether a virus will hop from species A to species B" there might be more data, since there are lots of species pairs.

"If one wanted to make a serious effort to forecast this, one can imagine a Monte Carlo model..." This is basically what the institute for progress did, but in a more simple linear way have you had a look at their calculations?

I haven't looked, could you provide a link?

Oh sorry I guess I missed it. I'll reply to that comment.

Re: the parvovirus, I've been chatting with Elizabeth on LW. She says that flu viruses tend to be unusually good at species-hopping, due to their ability to reassort with viruses native to the target species. Seems like the parvovirus doesn't have that ability, which could help explain why it's not a problem in humans.

Replying to the Institute for Progress analysis you linked.

Looks like this was published Feb 2023 and doesn't account for new developments since then? I don't see any discussion of mice or cattle, mammals which have more human contact than mink.

This sentence seems to underpin much of the IFP's optimism:

It is very difficult for R0 of a virus that is currently poorly adapted for human-to-human transmission to have a R0 that exceeds 1.0.

The CDC states:

Spread of H5N1 bird flu viruses from mammal to mammal is thought to be rare, but possible.

I'm not sure how to reconcile this claim with H5N1 being present in 20% of grocery store milk samples. I suppose shared milking machines could be creating exceptional circumstances.

It does seem like mammal-to-mammal spread gets you most of the distance from "pure bird flu" to "human flu", if I'm reading this article (from April 2023) correctly.

In any case, the IFP's R0 point seems a bit reassuring. It suggests we may get "warning shots" in the form of small-scale human-to-human transmission. It also suggests that much of our focus should be on reassortment: either a cow gets human flu, or a human with the flu gets H5N1.

Raw milk could also be a huge deal. Apparently it can have up to a billion virus particles per mL. Imagine someone drinks a liter of raw milk and consumes a trillion virus particles. How many of those are destroyed during digestion? What are the odds that one of those trillion particles will be viable virus with mutations that happen to make it well-adapted for humans? What if the person drinking the milk has an ordinary seasonal flu, creating the possibility for reassortment?

Yeah I just want to reiterate that I think this is the most prescient pandemic risk that the world has right now and I agree we should be investing a lot more in it than we are right now.

It's only the probabilities and as not of theory we disagree on I think, which probably doesn't change much in what we think should happen in practice.

I'm keeping an eye out for Sentinel's analyses: https://forecasting.substack.com/p/alert-minutes-for-week-172024

I'm worried too!

Thank you, that's very reassuring to hear. It can be hard to tell whether an issue is being overlooked or not

Thanks I appreciate the summary - better than anything else I could find on the internet, as is so often the case on this forum.

One small point is that there has been a long history of H5N1 infecting small numbers of humans, with limited human to human spread, without  the virus mutating to become highly contageous.   https://www.cdc.gov/flu/avianflu/h5n1-human-infections.htm. This suggests to me that the chances of the virus mutating to become highly contageous might be lower than is intuitive.

Also the fact that there are already 4 vaccines - although high rates of vaccine hesitancy may now unfortunately be established in the population.


I've spoken to some virologists who think the probability of it becoming a pandemic is low. However even tho there has been limited human spread for years, I am not aware of outbreaks rather than very limited infections among other mammals in past outbreaks like we are seeing now. I'm getting pretty worried, although I haven't yet come up with probabilistic predictions.

Also one thing I am not clear on is the efficacy of tamiflu or whether focusing on increasing tamiflu supplies might help.

Tamiflu is oseltamivir. CDC has info on antivirals and vaccines for H5N1 and H7N9.


Tl;dr: we have 3 antivirals, but some evidence of resistance, and a small stockpile of candidate vaccines, though the virus mutates fast.


Antiviral Drugs Can Be Used to Treat Illness

CDC currently recommends treatment with a neuraminidase inhibitor for human infection with avian influenza A viruses. Analyses of available avian influenza viruses circulating worldwide suggest that most viruses are susceptible to oseltamivir, peramivir, and zanamivir. However, some evidence of antiviral resistance has been reported in HPAI Asian lineage avian influenza A(H5N1) viruses (“Asian H5N1 viruses”) and Asian lineage avian influenza A(H7N9) viruses (“Asian H7N9 viruses”). Monitoring for antiviral resistance among avian influenza A viruses is crucial and ongoing.

CDC has posted guidance for health professionals and laboratorians.

U.S. Government Stockpiling Asian H5N1 and Asian H7N9 Vaccines If Needed

The United States federal government maintains a stockpile – vaccines, including vaccine against Asian H5N1 and Asian H7N9 viruses. The stockpiled vaccines could be used if similar viruses were to begin transmitting easily from person to person.  Since influenza viruses change, CDC continues to make candidate vaccine viruses as needed. Creating a candidate vaccine virus is the first step in producing a vaccine. More information about “Making a candidate Vaccine Virus (CVV) for a Highly Pathogenic Avian Influenza (Bird Flu) Virus” is available.

Thanks Monica! I suppose the big question is, do outbreaks among mammals make future outbreaks among humans more likely. Intuitively it might seem so, but  I can't think of other examples off the top of my head where there was a known and tracked outbreak of a non-human affecting disease among mammals, which then mutated to become a very contageous outbreak among humans. There may well be cases though I'm unaware of.

This is very different from diseases like the plague, or sleeping sickness where the host animals such as rats or cows spread the disease directly to humans.

Very keen to hear examples if anyone knows of them.

The main concern now is that we’re seeing unprecedented likely mammal-to-mammal spread. I don’t know what seems intuitive to most people, but this is the reason for heightened concern around the issue. Whatever your priors were before, this should probably update you towards a bigger risk of human-to-human spread.

Yes they do update me to a higher risk of human to human spread.  Human to human spread however has happened a number of times before, yet has never got out of hand. This pattern over the last 10 years. could perhaps update us towards a smaller risk of highly contageous spread between humans.

Of course we should take this seriously, and we may well not be taking it seriously enough right now. Knowing the history of a disease is important though, as this is not a new disease and is more likely to follow past trends than new noes.

Full agree! A track record of human-to-human spread with limited transmissibility is an update against a change in that area, and a new event of rapid mink-to-mink spread is an update for. Not sure how to balance them out mathematically, but it's important to keep both in mind.

Also, with regard to human vaccines, my understanding is they’re in low stock and only tested for seroconversion, since we don’t use them widely enough to get direct data on efficacy. We also have poultry vaccines but don’t use them widely, and it seems their effectiveness is inconsistent. So much to be improved on in the realm of H5N1 vaccine tech.

Changed thanks nice one!

Thanks for the good post, and raising a critically important issue. I absolutely recognise the individual disconnect between awareness and action at the beginning of the COVID pandemic, and think posts like this one are a good way to avoid this. Some things to think about:

  • I think Zeynep should be held in good regard as a source. I found her reporting throughout the COVID pandemic to be consistently excellent, readable, and epistemically robust.
  • The spread in the mink farm is another case where humanity's collective disregard for animals turns out to be collectively harmful, just like the wet markets in Wuhan. This seems to be a case where different EA branches are converging on a single clear policy direction. Seems like this could be a case where EA might be able to make policy progress in the wake of a near-miss event?
  • Even if the likelihood of this becoming a pandemic remains low, are governments reacting proportionally to the risk? Maybe they can't/don't want to say so publicly, but given the slow response to COVID my prior is that they're much more likely to be reactive rather than proactive.

I've gathered more information on the current status at my H5N1 pandemic prediction checklist.

I think claiming 10,000 lives but being nowhere next COVID scale is pretty unlikely. What mechanism do you suggest would control it at this point?

I don't think I suggested that? Forgive me if the original post was phrased poorly, I wrote it in some fourty minutes.

My point with mentioning 10,000 lives lost was to operationalize the question of whether it becomes a serious pandemic.

Apologies, I think the juxtaposition of "non-zero chance of >10,000 deaths" and "unlikely to be COVID scale" led me to think you were claiming something different

Agreed. Here's the account I would give, using my pandemic prediction checklist framework.

  • Currently, H5N1 scores a 5/14. COVID scored a 13/14.
  • +1 point: You'd have a death toll of > 2,000, 5x beyond when we saw the stock market crash during COVID
  • +1 point: Almost certainly, the disease would have altered to be transmitted efficiently from human to human via respiration. Human-to-human H5N1 will lead to a very serious response to contain it, and if we can't contain it at a few poultry/mink farmworkers, that's a sign it spreads very efficiently.
  • +1 point: That transmission mechanism would likely mean it has led to community spread
  • +1 point: Almost certainly this would be front page news because bird flu is scary even when it's mostly just in birds and animals and sporadically in farmworkers.
  • Possible +1 point: Once we get to 10,000 deaths, it's hard to imagine a scenario where that's true but it's contained to a single country. Could happen (a Chinese mink farm leads to efficient human transmission; a Wuhan-on-steroids lockdown isolates it perfectly from the rest of the world), but doesn't seem likely.
  • Possible +1 point: If it's limited to one country, they might choose to quarantine the region to contain it. Global spread may see a return to things like travel bans to island nations or other serious containment measures. Pretty easy to imagine a situation with at least one going on.
  • Possible +1 point: A 50% CFR from a respiratory infection could easily overwhelm hospitals
  • Possible +1 point: If we started to see all these problems, that would almost certainly motivate pharma to rush for mRNA vaccines and other tools

That puts it at anywhere from a 9/14 for the "almost certain to occur by 10,000 deaths" checklist items to a 13/14 (a score that's far more plausible to me).

I'm working with the assumption that we can either test for it sufficiently well, or that given how deadly it is, we just enforce zero tolerance for social contact for any symptoms of illness. If we had testing difficulties, failed to enforce strict social distancing, or had to deal with asymptomatic spread, then I think by the time we're at 10,000 deaths, we're probably already at a full-blown 14/14 situation that, if not already categorized as a pandemic, is well on its way.

I'm curating this — thanks a bunch for making it. I'd be very glad to see more comments and resource-sharing here, and more posts like this in the future in similar situations.  

Note also two recent posts from DirectedEvolution: 

One thing I'm curious about, given the predictions on whether the WHO will declare a PHEIC for H5N1: how come there wasn't a spike (or even growth) in markets like this one? The mink-to-mink transmission seems to be concerning, and that happened and was known in October. 

It could just be that reason the market isn´t moving is NOT because traders didn´t find it concerning but instead because the market is too low liquidity to be efficient. That´s very much my impression of low liquidity markets at least (I.e. there´s only 20 traders total in this case)

I’m not sure exactly when it became clear that we’re having a bad bird flu year, but it’s possible the mink farm outbreak wasn’t that big an update for the forecasters who were already seeing lots of transmission to mammals, just not among them.

(as of Feb 24) So far there’s a Cambodian girl that has already contracted H5N1 and died, others possibly detected. How does this update everyone? https://fortune.com/well/2023/02/23/h5n1-bird-flu-death-cambodian-girl-multiple-contacts-sick/

Other related questions on Metaculus:

Launches tomorrow:

Again this question can be misleading, as in the case of human to human transmission of this H1N1 (which may well happen, 25% chance seems about right to me based on past outbreaks), by far the most likely scenario is one or a few cases of transmission as we have seen in the past - not the kind of human to human transmission which could lead to disaster

I think it wouldn't be a problem if either the question had been phrased in terms of the expected number of human-to-human cases, or if there were other questions for specific orders of magnitude. I think the latter could still be done, and it would contextualize the first question.

Good points. As is often the case with forecasting questions, the details of the resolution criteria matter quite a bit. For this question to resolve as "yes",  the UK Health Security Agency must classify H5N1 as a "level 5 risk." Currently it is classified as a level 3 risk

So, forecasters on Metaculus are predicting whether the UK Health Security Agency will raise the H5N1 risk level to level 5, which signifies human-to-human transmission.

Zvi recently shared a post on H5N1 on LessWrong. 

Curated and popular this week
Relevant opportunities