Review of Tomas Pueyo's articles on COVID-19 management

This article by Dan Polansky looks at the role of Tomas Pueyo (also spelled Tomás Pueyo^[1], identified on Amazon as Tomas Pueyo Brochard^[2]) as relates to COVID-19 in 2020. It critically reviews his two key articles from a perspective of someone who is not an epidemiologist but rather a software engineer who reviewed and moderated fairly many engineering documents. A properly qualified reviewer could produce much better list of issues/defects.

This seems rather relevant since, the article "Why You Must Act Now" by Tomas Pueyo from 10 Mar 2020 had over 40 million views (believing the article page itself) and probably influenced politicians toward lockdowns. The article "Coronavirus: The Hammer and the Dance" by Tomas Pueyo from 19 Mar 2020 had over 10 million views (per the article page itself) and also stood a chance to influence Western policy makers. Pueyo has no credentials specifically for the data analysis he performed in the article, and one should analyze the risk that the lack of credentials and experience would lead to grave errors in his analysis. Also problematic was that his articles were not reviewed; generally, serious reviewing of artifacts greatly contributes to reduction of mistakes, to the extent that reviewers can sometimes be considered to be de facto co-authors.

Article: Why You Must Act Now

Issues

1.1) (Relatively minor) The article has no table of contents. That makes it harder to review whether the structure of headings is adequate for the analysis undertaken.

1.2) The article starts with something like an agitation/propaganda. This should perhaps send a warning signal that the material is low-grade.

1.3) Re: "Exhausted healthcare workers will break down. Some will die."

What is the quantification? Some? Many? At least two, per plural?

1.4) Re: "The only way to prevent this is social distancing today. Not tomorrow. Today."

Clearly untrue, on literal reading. This would mean that if countries do not implement state-enforced social distancing on the day of publication, they will not prevent "this". But surely the author cannot expect governing bodies to act on his advice within 24 hours.

1.5) Re: "Will I hurt the economy too much?"

The author pretends to understand worries of decision makers in power, by asking this question. But the question is not addressed in the rest of the article. Therefore, the question is asked and then implicitly dismissed.

1.6) Re: "This works extremely well when you’re prepared and you do it early on, and don’t need to grind your economy to a halt to make it happen."

The worry is not about grinding economy to a halt but rather badly damaging the economy. The employed rhetorical technique is of hyperbole. Moreover, the concern is not just with economy in the abstract but e.g. with those people who become unemployed.

1.7) Re: "But in 2–4 weeks, when the entire world is in lockdown, when the few precious days of social distancing you will have enabled will have saved lives, people won’t criticize you anymore: They will thank you for making the right decision."

This predicts that in 2-4 weeks, the entire world is going to be in lockdown, a strongly quantified statement that was unlikely to be true and that turned out to be untrue. As per W:COVID-19 lockdowns by country, "A few countries and territories did not use the strategy, including Japan, Belarus, Nicaragua, Sweden, South Korea, Hong Kong, Taiwan, Tanzania, Uruguay, two states in Brazil and certain United States states."
As for "people won’t criticize you anymore": which people? Surely some people are going to criticize the lockdown in part since they will have different priorities, risk aversions, biases, etc. And that really happened in 2020 and later: in many countries, there was a continuing debate concerning the lockdowns, with various parties sharply criticizing the lockdowns.

1.8) Re: "South Korea cases have exploded, but have you wondered why Japan, Taiwan, Singapore, Thailand or Hong Kong haven’t?[...]Taiwan didn’t even make it to this graph because it didn’t have the 50 cases threshold that I used. Many of them were hit by SARS in 2003, and all of them learned from it. They learned how viral and lethal it could be, so they knew to take it seriously. That’s why all of their graphs, despite starting to grow much earlier, still don’t look like exponentials."

The language of "knew to take it seriously" contains no observable or clearly testable predicate. It tells us very little.
Expanding on the above, there is no description of e.g. what exactly did Japan do to prevent the growth of cases.

1.9) The article does not contain the words/phrases "economic", "long-term", "liberty", "human right" and "overreach". It does not seem serious about any concern other than the immediate or near-future saving of lives. One could well conclude that certain economic concerns are to be overriden, but to do so, one would have to take economic concerns into account in the first place. Despite that, the beginning of the article indicates lockdowns to be "the right decision".

1.10) Expanding on the above, there are different political positions on what is right and important. There is Millian utilitarism with the greatest happiness of greatest numbers; there is libertarianism with its emphasis on liberty. The article makes claims about "the right decision" without acknowledging that it is making an unspoken assumption of what is ultimately important. The assumption that human death avoidance is the sole thing of ultimate importance is untenable since to achieve that, one can sterilize the Earth's population and make sure there are no longer any human deaths after a point in foreseeable future, but that is hardly an acceptable or attractive prospect.

1.11) Re: "The total number of cases grew exponentially until China contained it. But then, it leaked outside, and now it’s a pandemic that nobody can stop." [...] "But they aren’t enough to get us below 1 for a sustained period of time to stop the epidemic."

A contradiction: on one hand, nobody can stop the pandemic, on the other hand, governments are allegedly not doing enough to stop the epidemic and should do more.

1.12) Re: "Here’s what I’m going to cover in this article, with lots of charts, data and models with plenty of sources":

This suggests that "lots of charts, data and models with plenty of sources" is in itself something good enough to bring about validity. It isn't.

1.13) The article does not seriously reckon with any potential negative impacts of a lockdown. This was hinted at in 1.4); in addition to that, there is no mention of additional suicides, missed visits to medical doctors, etc. The article should have at least acknowledged existence of these items.

1.14) The article does not mention or refer to any existing epidemiological plans. If it did, it would need to explain why the existing plans and the pre-pandemic positions of epidemiological experts on the suitability/propriety of lockdowns were incorrect.

Impact

As per the article itself, the article got 40 million views in a certain week; and over 40 translations were created and attached to the article.

Endorsements

The article has attached a list of endorsements[1], featuring many notables.

Critics from comments

Selected critics from comment section in Medium:

Margaret Menzin, Professor of Computer Science and Mathematics and Program Director of Mathematics and Statistics
Bastian Schoell
Don Planck
L.S. Mech
Robert Dartt Webster

Article: The Hammer and the Dance

Issues

2.1) Re: "Countries have two options: either they fight it hard now, or they will suffer a massive epidemic. If they choose the epidemic, hundreds of thousands will die. In some countries, millions."

The specific quantitative meaning of the above is very unclear. "hundreds of thousands" are going to die where? In some countries? In most countries?

2.2) Re: "We will be locked in for weeks, not months."

It is unclear from which statements this conclusion follows. It was not borne by actual experience of the countries that did lockdown.

2.3) The article states that in "do nothing" scenario, 10 million people will die in the U.S., but it does not derive the figure in any way but rather defers the figure to "epidemic calculator" http://gabgoh.github.io/COVID/index.html. The calculator itself contains no data validating that calculator. The calculator states: "This calculator implements a classical infectious disease model — SEIR"; this is a simplistic model that, as far as I know, has never been shown to be numerically adequate to real-world development of epidemics.

Actual U.S. COVID deaths as for Jan 2024 reached over 1,100,000 per various sources. It seems unlikely that without lockdowns, the figure would have been 10 times worse.
In a proper exercise, Pueyo would have named the model as "SEIR". He should have pointed to sources establishing numerical adequacy of this model in relation to real world.
Moreover, Pueyo should have analyzed the credibility of http://gabgoh.github.io/COVID/index.html rather than taking it automatically to be a correct implementation of a model.

2.4) Re: "So why is the fatality rate close to 4%? If 5% of your cases require intensive care and you can’t provide it, most of those people die. As simple as that."

Nothing is simple or obvious about it. It is not clear what "most" refers to quantitatively, how exactly was the input figure 5% obtained and then where the factor reducing 5% to 4% came from. This is not scientific or science-like writing, nor serious engineering writing.

2.5) Chart 7 indicates the figure under "suppression" strategy to be 4000 (four thousand).

The figure is probably for the U.S. Since the chart does not clearly state "for the U.S." or the like, it could be dismissed purely on formal grounds.
The figure is confirmed in the statement "Under a suppression strategy, after the first wave is done, the death toll is in the thousands, and not in the millions."
It is hard to believe that the death figure for the U.S. could have been so low even under some radically tightly executed suppression strategy.
Even in Taiwan noted for its early successful suppression, as of Jan 2024 there are 19 000 COVID confirmed deaths, given population of over 23 million.

2.6) Re: "Unbridled coronavirus means healthcare system collapse, and that means mass death."

The phrases "healthcare system collapse" and "mass death" are likely to be free from any specific meaning useful for a serious analysis.
The term "collapse" would, in a house, indicate elimination of its residential capacity. An analogue in healthcare would be a drop to healthcare performance to zero, but that is obviously not meant. What is meant is a reduced ability to serve all the needs that would be served otherwise. To use the vivid verbal image of "collapse" for something that is nothing like that is not serious.
The virus is nearly certainly going to cause "mass death", as does influenza, as long as the deaths are at least in the thousands.
The above statement is in a large font as if it was some kind of important takeaway, but this is 1) inappropriate for a scientific article, 2) as just pointed out, there is almost no takeaway but rather empty rhetoric.

2.7) Re: "Put in another way: the mitigation strategy not only assumes millions of deaths for a country like the US or the UK."

Now we also get millions of deaths for the UK, which means at least 2,000,000 (given the plural). The actual death count as of Jan 2024 is over 230,000.

2.8) Re: "So once we’re done with a few million deaths, we could be ready for a few million more — every year. This corona virus could become a recurring fact of life, like the flu, but many times deadlier."

Incredibly, we are told that because the virus is going to be highly mutable, there could be a million deaths per year, going forward indefinitely. But downwards, he states "until we have a vaccine", so then, not really indefinitely.
More on a point down below: If the virus is so highly mutable, as he says, are the vaccine makers going to be able to keep up with the new mutations?

2.9) Re: "The US (and presumably the UK) are about to go to war without armor."

Metaphors like this are a tool of political rhetoric, not science and engineering.

2.10) Re: "Wouldn’t it be dumb to commit to a strategy that throws us instead, unprepared, into the jaws of our enemy?"

More of empty rhetoric.

2.11) Re: "On the other, countries can fight. They can lock down for a few weeks to buy us time, create an educated action plan, and control this virus until we have a vaccine."

It seems unlikely that one could have known that the "suppression" as opposed to "mitigation" lockdown would require only "a few weeks"; and in most countries that went for lockdown, it did not.

2.12) Re: "What if Churchill had said the same thing? “Nazis are already everywhere in Europe. We can’t fight them. Let’s just give up.” This is what many governments around the world are doing today. They’re not giving you a chance to fight this. You have to demand it."

More empty rhetoric with no added informative value.

2.13) There is no mention in the article that part of the confirmed case growth is due to the growth of available and performed tests.

2.14) Re: "Here’s what we’re going to cover today, again with lots of charts, data and models with plenty of sources"

As with the other article, lots of charts, data and models do not establish validity.
The article does not really present "models", their analysis and validation; for instance, it derives its 10 million deaths figure with the use of model that it does not name and whose mathematical formula it does not state, merely linking to http://gabgoh.github.io/COVID/index.html. (A similar defect is in the first article). The page http://gabgoh.github.io/COVID/index.html states this: "The clinical dynamics in this model are an elaboration on SEIR that simulates the disease's progression at a higher resolution, subdividing I,RI,R into mild (patients who recover without the need for hospitalization), moderate (patients who require hospitalization but survive) and fatal (patients who require hospitalization and do not survive). Each of these variables follows its own trajectory to the final outcome, and the sum of these compartments add up to the values predicted by SEIR. Please refer to the source code for details. Note that we assume, for simplicity, that all fatalities come from hospitals, and that all fatal cases are admitted to hospitals immediately after the infectious period." Directing a reader to the source code is not a proper way to specify a model for reviewers; in a properly conducted exercise, the calculation method/detail has to be available in English. Moreover, the page does not seem to link to the source code; from using Google and browsing around, one can guess that the source code is at https://github.com/gabgoh/epcalc, but the page does not seem to explicitly link there. The author of the page is Gabriel Goh[2], currently "machine learning researcher at OpenAI".

Overall, Tomas Pueyo cannot know all these things he is putting forth with so much confidence, and so much urgency. The article does not provide proper traceability for most of its key statements: we do not know which statement is supported by which other statements.

Could the public know the article was wrong? Public could have used its strategy of relying on properly published peer-reviewed science. People could have thought: I do not have enough expertise in reviewing epidemiological articles to find whether the modeling and calculation it uses is completely wrong. But even that is probably not quite true: even a surface review as the one above could have dismissed the article mostly on relatively easily accessible grounds.

Instead of sounding an alarm about the article, many people endorsed the article, some collaborated on it, and some have translated it into other languages.

Why did the article need to be translated into other languages? Would it have been because political decision makers do not know English enough? A more likely explanation is that the article was addressed to the general public, which should have put pressure on politicians to adopt a suppression lockdown strategy. The tone of the article seems to be this: you, the general public, can see that the article is written with lots of data, charts and models and as you can see, its conclusions are a no brainer. You should be convinced by the superficial persuasion techniques; you do not need to be worried that you do not have the expertise to judge the matter.

Endorsements

The article links to a list of endorsements[3]. The list feature Tim Berners-Lee, Steven Pinker, and Daniel Dennett. And yet, arguably, the article should have been dismissed on relatively easily accessible grounds.

Article: The Fail West

This article is covered here as something of an aside: it was not part of the two key 2020 articles with millions of viewers that probably influenced Western policy makers.

Link: https://unchartedterritories.tomaspueyo.com/p/the-fail-west; 6 May 2021

Pueyo compares the lives lost in the West with those in Asia-Pacific.

Start of the article:

"Soon, over 1.5 million people will have died of COVID in Western countries.

"1.5 million futile, needless deaths. 1.5 million wasted lives.

"Meanwhile, in a block of Asia-Pacific countries with a population over twice as big, they lost 18,000 people."

Issues

3.1) In the first half of 2021, the pandemic was far from over. It was not the time on which to do the reckoning/accounting. This is especially obvious in retrospect since we know China had its late Covid wave much later, in 2022 per Wikipedia: COVID-19 pandemic in mainland China. But it was also obvious in 2021 since it was already in 2020 that Anders Tegnell pointed out that Covid needs to be managed in the long run rather than in mere several weeks or months.

3.2) It is not clear that data collection in Asia-Pacific is comparable in its thoroughness with the West. In particular, the data from China could back in 2021 be safely assumed to be completely wrong. Even today in 2024, we do not know the covid-positive deaths from China (or do we?) and we do not have all-cause deaths for China (or do we?). By contrast, we have solid all-cause death data for the West. Thanks to Karlinsky & Kobak^[3], we have all-cause mortality also from other parts of the world, including South America and Russia, but it is Asia-Pacific that is conspicuously absent from the coverage of that data set.

3.3) The folowing pleonastic statement is pure empty rhetoric or theatricals: "1.5 million futile, needless deaths. 1.5 million wasted lives.". One could have as well said "1.5 million lives needlessly lost" and be done with it. And a life lost is not automatically a live that is futile or wasted.

A more thorough review is pending; the above issues were found in quick first-impression review.

Responsibility

In the subject of responsibility of intellectuals, one can ask to what degree and in what way can the author be held responsible/accountable for this kind of articles, which, given a calm analysis available in retrospect, easily fits into the "spreading misinformation" category.

On a tangential yet somewhat related note, those eager to censor misinformation are all too likely to censor information and let misinformation flourish. But that is for a separate debate.

Online criticism

Some criticism can be found in the comment sections of the articles. In so far as the author did not post replies to these comments there, he failed to address them. Should I take an imaginary role of a review moderator, I would need to prevent closure of the review since not all comments have been addressed (accepted, rejected with an explanation, etc.). The documents would need to be put into the state of review still open, not all comments processed, not ready for publication or consumption.

A 2023 critical review is Senger 2023, in further reading.

Mainstream media

According to fortune.com, 'Donald McNeil, the eminent health writer for The New York Times, referenced “The Hammer and the Dance.” CNN’s Anderson Cooper invited Pueyo to speak on TV.'^[4]

One should perhaps not be surprised to find that mainstream media authors speak positively of things they are not qualified to properly critically evaluate. This is corroborated e.g. by the laudatory mainstream media response to lobotomy.

Conflict of interest

One critic pointed out that Tomas Pueyo, a stakeholder in a company publishing online courses, stands to benefit from various countries going into lockdown and the demand for these courses increasing. Could Pueyo's psyche unconsciously contribute to the mistakes and biases in his article? It seems hard to believe, but could human psyches be like that?

Personal note

I sent a link to Tomas Pueyo's "Coronavirus: Why You Must Act Now" to a family member via email. I am not beyond reproach. We are all fallible.

References

↑ Tomás Pueyo - Uncharted Territories, fr.linkedin.com
↑ Tomas Pueyo Brochard, amazon.com
↑ https://github.com/akarlinsky/world_mortality
↑ The overnight coronavirus expert by Adam Lashinsky and David Z. Morris, 10 Aug 2020, fortune.com

Review of Tomas Pueyo's articles on COVID-19 management

Contents

Article: Why You Must Act Now

Issues

Impact

Endorsements

Critics from comments

Article: The Hammer and the Dance

Issues

Endorsements

Article: The Fail West

Issues

Responsibility

Online criticism

Mainstream media

Conflict of interest

Personal note

References

Further reading