The War on Data: Now we play the price

A few years ago, Mark Palko wrote an article, The War on Data, where we wrote:

The value of shared data reaches its logical extreme in high-quality, publicly available databases such as those maintained by the U.S. Census Bureau. These sources do not just support an extraordinary amount of research; they help individuals and institutions make better decisions and give us a set of agreed-upon facts that help keep our discussion honest and productive. For all these reasons, recent threats to publicly available data are cause for concern.

I’d forgotten about this until I came across a series of recent posts from Palko reprising the themes. This is all particularly relevant now, when we have the oddity of super-precise reports of the stock market and unemployment filings but no comprehensive sampling for coronavirus testing.

Yasmeen Abutaleb, Josh Dawsey, Ellen Nakashima, and Greg Miller tell the story in this news article:

The United States will likely go down as the country that was supposedly best prepared to fight a pandemic but ended up catastrophically overmatched by the novel coronavirus, sustaining heavier casualties than any other nation.

It did not have to happen this way. Though not perfectly prepared, the United States had more expertise, resources, plans and epidemiological experience than dozens of countries that ultimately fared far better in fending off the virus. . . .

The failure has echoes of the period leading up to 9/11: Warnings were sounded, including at the highest levels of government, but the president was deaf to them until the enemy had already struck.

The Trump administration received its first formal notification of the outbreak of the coronavirus in China on Jan. 3. Within days, U.S. spy agencies were signaling the seriousness of the threat . . .

The most consequential failure involved a breakdown in efforts to develop a diagnostic test that could be mass produced and distributed across the United States, enabling agencies to map early outbreaks of the disease, and impose quarantine measures to contain them. . . .

It does seem like a failure in not recognizing the value of good public data.

66 thoughts on “The War on Data: Now we play the price

  1. I think the US is a pretty big success story. It wasn’t until it hit the US that “intubate early” was ended due to a critical care doctor who was forced to step down from the ICU and turn to social media over it. In other countries he would have been jailed for going against authority like that. Many deaths from SARS were probably due to the same policy.

    Then again the FBI is raiding vitamin c clinics and FCC sending letters to HBOT clinics despite those being the two most logical and promising treatments while billions were thrown at the expensive and questionable approaches of ventilators and now a vaccine. And still no one has published a study checking for ADE where it was seen for SARS, so there is lots of low hanging fruit.

    • Also, the tobacco smoking and asthma connection to high altitude sickness and covid is another thing that has gotten some attention, but way less than it should have since people are so biased against smoking (I really think many would prefer civilization collapse than admit a benefit to smoking). That is another low hanging fruit the US has been dropping the ball on.

      But the ending of intubate early (that there was never any evidence for, yet was recommended by the WHO) was such a huge step forward that probably cut down mortality rates by 5-10x.

      • Anoneuoid,
        I don’t see how the U.S. can be reasonably considered to be a “pretty big success story.” In deaths per capita we are among the worst in the developed world, and it looks like we are going to climb even higher up the list over the next month.

        As to smoking: I agree with you that potential protective effects of smoking have been understated, and probably under-studied. I even agree that this is probably due to ‘bias’, depending on how you define bias. But that ‘bias’ is pretty well-founded: the tobacco industry spent decades lying about the effects of smoking, so people are suspicious about any claim that smoking is beneficial in any way. Also, smoking has proven deleterious effects on lung function, so the idea that it could reduce susceptibility to a disease that infects and attacks the lungs is counterintuitive. Basically I think people have very good reason to be suspicious of claims that smoking is beneficial when it comes to COVID-19…but this does not mean that they should completely dismiss the possibility of a benefit.

        • I asked for this per capita death rate data below but didnt get an answer. Theres some discussion by others of what I was going to check, but im not sure if thats a good number to compare. Even over time in the US the testing process is changing. I also think nursing home deaths should probably be looked at seperately.

        • > I asked for this per capita death rate data below but didnt get an answer.

          The ball is in your court though. You keep asking people to prove to you why covid19 has been bad in the US, but we’re asking you to prove why it’s been good.

          You offer a heartwarming story of a scrappy ICU doctor who risked his career to save us from ventilating ourselves to death, but that isn’t enough. I’m certainly not tied in to ICU procedures enough to even know how to evaluate that statement and weigh it against all the other covid impacts.

          Indeed, the other stories you tell are:

          > Then again the FBI is raiding vitamin c clinics FCC sending letters to HBOT clinics despite those being the two most logical and promising treatments

          > while billions were thrown at the expensive and questionable approaches of ventilators and now a vaccine.

          > And still no one has published a study checking for ADE where it was seen for SARS, so there is lots of low hanging fruit.

          So doesn’t that mean the US response has been, by your own definitions, bad?

          And like lemme zoom in on that a bit:

          > now a vaccine

          Huh? Are you on some anti-vaxxer stuff? I talked to someone a little under a month ago about covid and they were bemoaning that we spend all our time researching vaccines for single diseases and it’s silly — why aren’t we researching a cure for all viruses! Indeed, they said the FDA had banned some sort of vitamin B thing that actually was the cure for everything (I believe this included AIDS).

          Anyway, I don’t get it. I don’t know much about drugs and stuff. Maybe with the right mix of vitamin C and vitamin B counterfactual me could live to 200. I don’t exclude this possibility, but I do find it genuinely difficult to seriously engage with your arguments partly because I have to give up on a lot of my other understanding of the world (vaccines good).

      • Are you saying my definition excludes those things? Most excess deaths were caused by treating people in the most dangerous way possible (putting people on ventilators) and sending infected people into nursing homes. I’m saying the US did contribute to ending the first thing, and now mortality rates are much lower. That was huge.

        I don’t know why people were sent into nursing homes but obviously they should have been locked down same as for influenza. That was a huge failure that shouldn’t be repeated.

        I doubt testing everyone, wearing masks and closing businesses will be a net benefit at this point though. In Jan/Feb it would have made sense, but back then there were touchscreen robots going around NYC spreading it, etc and the media saying not to care.

        • > Are you saying my definition excludes those things?

          Yes. The U.S. has more excess deaths and a longer economic disruption than most of its peers, and you’re calling it a big success story. It’s possible that great work has been done in the U.S. to prevent excess mortality in the future, but the fact is that the past up to the present looks pretty grim.

          > I doubt testing everyone, wearing masks and closing businesses will be a net benefit at this point though. In Jan/Feb it would have made sense, but back then there were touchscreen robots going around NYC spreading it, etc and the media saying not to care.

          It seems like you’ve pretty consistently been of the opinion that R0 is buoyed up in large part by these relatively isolated incidents like your touch screen example, claiming that “In every country with a lot of cases people have been spreading it on purpose”. It seems like you’re suggesting that as long as we avoid those incidents we’ll be fine. What do you think is happening in Florida right now?

        • The U.S. has more excess deaths and a longer economic disruption than most of its peers

          I’d have to see the data you are basing this off of.

          It seems like you’ve pretty consistently been of the opinion that R0 is buoyed up in large part by these relatively isolated incidents like your touch screen example, claiming that “In every country with a lot of cases people have been spreading it on purpose”.

          Yes, for both SARS and SARS2 it seems the vast majority of people don’t transmit to anyone or only one person, then a few transmit to many. Obviously if you do it on purpose it is more likely you become a superspreader.

          https://pubmed.ncbi.nlm.nih.gov/15030693/
          https://www.nature.com/articles/nature04153
          https://www.thelancet.com/journals/laninf/article/PIIS1473-3099(20)30287-5/fulltext
          https://wellcomeopenresearch.org/articles/5-67
          https://www.sciencemag.org/news/2020/05/why-do-some-covid-19-patients-infect-many-others-whereas-most-don-t-spread-virus-all

        • The US does not have more deaths *per capita* than “most of its peers”, however. The US is pretty firmly in the middle of the pack among developed relatively-large “Western” nations; below Spain, Italy, and the UK; above Germany, Canada, and Australia.

          The difference is that the US has a much higher rate of *current* deaths than those other countries.

          And the US really has no actual “peers” in this sense. The US has much more internal complexity (50 states, with a much greater urban/rural variation and cultural differences) and is far larger than any European nation, but has much more per-capita wealth than other large internally-complex nations with a federal structure (Mexico, Brazil, India).

          There are also uncertainties in the counting of deaths & differences between nations. I am not sure a, say, 25-30% difference either way is outside the margin of error. (Brazil and Mexico might have more per-capita deaths than US depending on the degree of undercounting, for example.)

          Iceland, Australia, and New Zealand have clearly done much better — but they have inherent geographical advantages. Germany seems to have done anomalously well compared to any other non-island “Western” nation.

        • confused –

          > The US does not have more deaths *per capita* than “most of its peers”, however. The US is pretty firmly in the middle of the pack among developed relatively-large “Western” nations;

          The US ranks as 9th worst in deaths per capita among over 200 countries. What are the reasons why you’d select “relatively-large, Western nations” as your comparison group? Why would being relatively large or Western be explanatory for the effectiveness or lack thereof of policies adopted?

          The US has a huge amount of resources to bring to the table, low population density and low use of public transportation relative to many of the countries you left out of the grouping.

          On the other hand, the US had a lot of international travel and a population that resists simple practices like wearing masks because so many American snowflakes think wearing a mask is a huge infringement of their rights.

          Cross-country comparisons are tough. Lots o’ variables that are hard to control. But I would say that if you’re going to to a grouping like you did, you should lay out the reasons why you made that grouping. Otherwise, it may be that you form the comparison group essentially on arbitrary criteria for the sake of confirming a bias.

        • >>What are the reasons why you’d select “relatively-large, Western nations” as your comparison group? Why would being relatively large or Western be explanatory for the effectiveness or lack thereof of policies adopted?

          Because the numbers are likely not trustworthy for very poor or unstable nations (due to deficiencies in testing/death reporting, and in some totalitarian cases probably also dishonesty). Small size also makes the numbers fairly meaningless — you could get very lucky, or one hotspot could overwhelm you. Size… basically means the numbers mean something, it’s not just pure luck or extreme isolation, or the effect of one hotspot overwhelming the nation’s numbers. The worst nation in the world by per-capita death rate (San Marino) is tiny, and some tiny Pacific islands have zero cases.

          So IMO the only nations that we have meaningful data on are large nations with a certain degree of wealth & stability. That’s largely the Anglosphere, Europe, East/South Asia, and the wealthier/more stable nations of Latin America.

          Among that group, I don’t think East Asia is a good comparison because the kind of strict measures that were possible in e.g. China are just not possible in Western democracies. Even democratic nations in East Asia have a far greater degree of social cohesion.

        • But you left off many countries that likely have reliable data:

          Switzerland, Poland, the Czech Republic, Portugal, Finland, Denmark, Norway, Ireland, Scotland, the Netherlands, Austria….

          > Among that group, I don’t think East Asia is a good comparison because the kind of strict measures that were possible in e.g. China are just not possible in Western democracies.

          But those are part of the relevant analysis!

          > Even democratic nations in East Asia have a far greater degree of social cohesion.

          Exactly!

        • It feels like you’ve defined the peer group so as to place the United States in the middle of the pack. I’m defining the peer group in terms of resources to play with, so roughly by GDP per capita. Not including Switzerland, Finland, Denmark, Norway, South Korean, Taiwan is a mistake in that regard. You’ve also essentially excluded the East Asian countries from the comparison because they had a well-organized response, which is absolutely baffling reasoning.

        • I can see your point. I guess it depends on *why* you are making the comparison.

          I didn’t think East Asia was relevant because, well, obviously those nations are fundamentally better set up to deal with this specific kind of problem. Just as New Zealand and Iceland are obviously advantaged because of geography.

          I was more thinking of the question as “did the US do better/worse than would be expected by comparing to roughly comparable nations” or “how much did specific bad decisions made during the pandemic or by the current administration hurt our response”. So for that comparison fundamental, long-standing differences in governmental structure are not relevant.

          Overall differences make a huge difference, sure! But they aren’t a useful guide to making any kind of decision. You just can’t turn the US into South Korea or Japan.

        • And I was absolutely including most of those European nations in the comparison (though not the former Soviet ones, and Scotland is part of the UK).

          I didn’t mention Ireland because IMO their per-capita death numbers are close enough to the US’s that differences in counting between nations mean we probably can’t confidently say which is higher. From https://coronavirus.jhu.edu/data/mortality – Netherlands 35, Ireland 36, US 42, France 45. I doubt the margin of error on the data is small enough to rank these nations confidently.

        • It doesn’t make sense to compare per capita death rates because most European countries are in a very different place in their pandemic trajectory than the USA. Spain is down to <1 daily death per 100,000 and stationary, Italy is about 3 per 100,000 and decreasing while the USA is about 30 per 100,000 and rising.

          Their total per capita death rates may be broadly similar but it doesn't seem likely to last much longer.

        • Yes, predicting future death totals is far more uncertain.

          But I think that is a bit pessimistic. The US doesn’t really have one pandemic curve, it has several in different regions. The first peak and decline were driven by the Northeast and Upper Midwest and the current rise is driven by the Southwest (CA/AZ/TX) and some Southern states (largely but not entirely FL).

          Hospitalizations seem to have stopped rising or nearly so in TX and AZ (*knocks on wood*). So I don’t expect continued exponential growth in deaths. The next 2-3 weeks are going to have high deaths from the high new-daily hospitalizations in June/first week of July but then things will likely improve.

          There are several other states that look bad and are likely earlier in their curve than AZ/TX/CA/FL, but are too small to really affect the national numbers much. (AZ+TX+CA+FL are something like 30% of US population.)

          Now even after this “peak” I am sure death rates will remain higher than current rates in Spain/Italy. But I expect a reasonably effective vaccine relatively soon, so I don’t expect those death rates to continue for all *that* long.

        • Florida, Texas, Arizona, etc. are cases of overconfidence; since they’ve been predicted to be “the next Italy”/a huge disaster since Spring Break (mid-March), and that didn’t materialize for 2 months, so by late May people kind of assumed that it was basically a Northeastern US/Western Europe problem and stopped being careful.

          Texas is now being much more careful (at least in the large cities) and things are stabilizing. I live in TX, but I imagine AZ and FL are similar.

          It’s interesting that California is also going up despite being hugely politically/culturally different and having different government responses from the rest of these states. There is an odd geographical pattern in the US (started out bad in the Northeast/Upper Midwest, now moved to the South/Southwest). But there are exceptions to even that — Georgia reopened first and yet deaths are not rising there as they are in TX & AZ.

        • confused –

          > Florida, Texas, Arizona, etc. are cases of overconfidence; since they’ve been predicted to be “the next Italy”/a huge disaster since Spring Break (mid-March), and that didn’t materialize for 2 months, so by late May people kind of assumed that it was basically a Northeastern US/Western Europe problem and stopped being careful.

          Again – that looks like highly selective reasoning to me. You jump into firmly determining the causes of outcomes based on simplifying complex psychological and emotional processes, and completely leave off any factors that might run in the other direction – such as a political orientation that leads to resistance to shelter in place policies and mask-wearing – as if you know for a fact that has zero explanatory power.

          If I”m not mistaken, a while ago you were looking at lower rates of transmission in some states and confidently attributing causality to seasonality – when now the potential impact of seasonality looks extremely complicated.

          > But there are exceptions to even that — Georgia reopened first and yet deaths are not rising there as they are in TX & AZ.

          Looks to me that you are over-fitting to one potential influence without considering enough context. Are there any factors that might help explain why “opening up” early in one state might have a different impact than “opening up” early in another? I certainly don’t know. Do you? If not, then you can’t eliminate “opening up early” as partially explanatory for spikes in Texas or Florida even if another state that opened up early hasn’t (yet) experienced the same kind of spike.

          From reading your comments of the last few months, I have a feeling that you should be somewhat more circumspect about drawing conclusions. I could be wrong about that. And at some level, speculation is fine. But even there, people need to be really careful when the data and evidence are so preliminary. Much of your reasoning is based on your anecdotal observations. Anecdotal observations are fine – but you need to check carefully for potential biases.

        • The relative lack of seasonality in the US data is very strange. Makes me think the numbers are mostly driven by testing policies. Maybe its something else…

        • I sounded far more confident than I actually am, sorry.

          >>and completely leave off any factors that might run in the other direction – such as a political orientation that leads to resistance to shelter in place policies and mask-wearing – as if you know for a fact that has zero explanatory power.

          No, my point is that MT has more governmental measures in place than SD, and TX has more measures in place than GA. MT’s numbers may be too small to be very meaningful, but hospitalizations have risen sharply there vs. overall fallen slowly in SD. TX has seen a sharp rise in deaths and GA a continued plateau in deaths.

          (Sure, TX has less measures than many other states, but it’s not the lowest.)

          % of population infected may also be a factor — SD and GA have more per-capita deaths than MT and TX — but it seems very unlikely that it could possibly be high enough in SD and GA to affect R *that* much.

        • >> If not, then you can’t eliminate “opening up early” as partially explanatory for spikes in Texas or Florida even if another state that opened up early hasn’t (yet) experienced the same kind of spike.

          Oh, no, absolutely I think opening up early was a causal factor for spikes in TX and FL.

          But I think it took *both* opening up early (as a governmental measure) and “overconfidence” – fairly widespread belief among the population that this was over/never relevant to TX and FL in the first place. Actual behavior hasn’t necessarily been strongly correlated with state-level measures.

          But the geographical trend still seems somewhat unexplained.

    • > I think the US is a pretty big success story.

      It’s too early to really evaluate, but failures in testing and contact tracing and isolating seem to me to be incredibly important indicators that there has been no “pretty big success story.”

        • Let’s assume that your argument is correct (I’m skeptical, but for the sake of argument)…

          Even if it’s true that in the US doctors changed the existing protocol to one which has led to much better outcomes, that doesn’t eliminate the potential of much greater illness and many greater deaths, differentially, caused by our failures on testing, tracing, and isolating relative to many other countries.

          So how would you way “success” in one area against poor performance in another?

        • weigh not “way.”

          It’s always amazing to me how I substitute homophones when typing. I think it’s because of some rather uniquely crossed wiring in my brain.

        • Joshua –
          Our brains are wired for spoken language.

          Written language is an attempt to represent spoken language without having to actually speak.

          If we were starting from scratch, way and weigh would be spelled the same, or would be represented by marks that aren’t meant to be read phonetically.

          So don’t blame yourself, blame your ancestors who casually let “way” and “weigh” come to sound the same. You’re just trying to get back to the roots of written language. An admirable but ultimately futile effort.

          Today’s language trivia item: look up the original meaning and pronunciation of “fomites”. It blew me away when I did that a couple of days ago.

        • John –

          I’ll check that out.

          Here’s what’s interesting about my error to me. I don’t think that I would make a similar error if I were writing by hand, or at least I would be less likely to do so. So there is something about typing a comment online which is more similar to speaking than writing by hand. I think that there is some connection to typing, and posting comments online, that is quicker and less deliberate (in part becsuee it is to edit so there is less “cost” in making a mistake) and inherently more “conversational.”

          Still – why, when I know that I meant “weigh” rather than “way” when I’m conceptualizing the sentence would I type “wrong” word when I type the sentence out?

        • What is going on that I would make a “mistake” from a mode based only on sound when UK not even making any sounds? It definitely seems to me that there’s a brain organization aspect of functions that reside in neighboring areas getting crossed up.

          I suspect that I make that error type more frequently than most. But then again, I also seem to make grammatical and sentence pacing errors more frequently than most. I think it’s because I tend to spend less time on editorial functions than most when I write comments online – and have a habit of not proofing carefully if at all. Maybe my error rate of that type happens more often just as the other proofing errors I make happen more often – but my sense is that there is something fairly uniquely prevalent about my substitution rate of homophones when I type online comments. I don’t blame myself, it’s just an observation.

          An intersting (to me) side note:I so an exercise with non*native speakers of English where I read them short passages of a text and have them say that passage back to me in a way that exactly preproduces what I read – including intonation and pitch and rhythm as well as the exact words (my idea is that the inconsistencies in what they producenfeom what I said will illustrate the patterns of their “errors” or native language structures that ehy can learn from). What’s fascinating is that in real time of repeating what I said, people can often substitute synonyms (words people more typically retrieve from their vocabulary bank) for the words in the original passage. I’m amazed by the brain power it takes to go out and retrieve a synonym within the microseconds between one word and the next.

        • It was a big success that I don’t think would necessarily be possible in other cultures. In china he probably would have been jailed, but many other cultures are more regimented. Even most of the US medical industry is pretty regimented, but we still managed to do it.

          Everything comes with tradeoffs.

      • I’d say the US *governmental response* has been very poor, but other factors such as healthcare improvements; relatively low population density of much of the US relative to Europe; and people and companies taking measures not required by government have made our outcomes less bad than would be predicted by the governmental response (relatively mid-range among ‘Western’ nations, though much worse than East Asia).

        IE – I don’t think it would have been expected in mid-March that by mid-July the US would have fewer per-capita deaths than Italy or Spain.

    • Excellent question. And I really can’t quite figure out what’s going on here.

      My instantaneous reaction was to assume that the data were now going to be diverted to a politically pliant division of HHS which would proceed to either suppress them, or falsify and disseminate them for propaganda purposes. That would be my strong prior about what this administration would do.

      But looking into it a bit further, the reason given, namely to speed up the data collection process, may in fact be true (though it wouldn’t necessarily preclude also suppressing or falsifying the data). It appears that in a meeting between Dr. Birx (part of the White House task force on coronavirus) and hospitals where she complained that the data coming in were not sufficiently timely, the hospitals pushed back and complained about the CDC’s data collection procedures and software. It appears that the resolution was an agreement to transfer the data collection to a private business that many hospital systems use to collect, manage, and store their own data. (The contract to TeleTracking was then set up with a no-bid contract, but I have been unable to ascertain whether the owner of that business has any political or other connections to the administration.) And the CDC is still receiving the data _from TeleTracking_. But it is unclear if it will be making it available to the public (so far, it isn’t), nor is it clear whether they are getting the raw data or some kind of processed version. Apparently nobody with authority to speak on the record about this is saying much.

      Moreover, it seems that attempts to suppress or falsify this data would fail in a public way because this data is also reported separately to Johns Hopkins and other sources that, in turn, do make the data publicly available. While this administration doesn’t hesitate to prevaricate even when the facts are readily available elsewhere, it sees like this would be a particularly clumsy way for them to proceed when they can just leave the data at the CDC and lie about it in their messaging and most people wouldn’t be any the wiser.

      Finally, there is the dog that didn’t bark. Democratic politicians have not been reticent to call out the administration for deception and corruption. Yet virtually none have spoken up about this at all. This is certainly something that could be used to stoke more public outrage than arcane dealings in the Ukraine. But nobody is speaking up. It makes me think that those with access to the actual situation perceive that nothing duplicitous is going on.

      Then again, I’m not very trusting of Democrats either. So I don’t know, but I’ve certainly moved off my prior to a more agnostic position about what’s going on here.

      • Actually, just saw that congressional democrats have, indeed, cried foul over the new reporting rules.

        That said, I appreciate the background you provided, since it suggests that the rule change might have some legitimate reasons.

        Either way, it still seems pretty stupid to roll out such a change on short notice in the middle of an ongoing crisis. It seems to be an example of a corollary of Clarke’s Law: any sufficiently incompetent governance is indistinguishable from fraud.

    • The cdc covid data has been crap this whole time. Do people still look at it? Use covidtracking.com that gets data directly from the states. Still has tons of issues but much better than the cdc.

      All cause mortality data from the cdc seems to be best though.

        • The CDC data is slow, but will be the best for the long-term picture of mortality impacts (number of excess deaths, distribution of deaths by age, etc.)

          But it’s not very good for a real-time picture at all…

        • That’s been my conclusion. Does anyone know of a good database that collects cases by age? I know not all states report this (I don’t know why) but is anyone gathering this together in one spot?

  2. Well if you want to talk about big screw-ups that lead to a lot of mortality and economic damage, let’s not forget:

    1) the CDC specifically advised to people **NOT** wear masks;
    2) the WHO is **STILL** resisting the obvious – that floating aerosols are the primary mechanism of transmission – even though some people recognized this before the first case occurred in the US.
    3) a variety of organizations were flogging all kinds of nutcake analyses and making wild and unsupported predictions from them to guide policy, all of which had pretty much zero positive impact despite the hours of chin-rubbing they generated, and created a major distraction from the relevant information;

    All of this misinformation contributed to major economic damage and hardship and almost certainly to substantial spread of the infection and higher mortality.

    So, sure, it would have been great to have an army of people tracking down cases and throwing out Q orders. But the fact is that even much of that could have been prevented if CDC and many other “experts” hadn’t been so stuck on a pre-determined idea about the transmission mechanism and completely oblivious to the accumulating evidence of asymptomatic transmission.

    • > 2) the WHO is **STILL** resisting the obvious – that floating aerosols are the primary mechanism of transmission

      I’ve read the letter from the aerosol experts, and they invoke the precautionary principle: this means that data on aerosol transmission is actually unclear. I’ve also dug for data, and there isn’t much. There are models on the possibility of aerosol transmission, but at this point what’s needed is hard, experimental evidence.

      • Needed for what though? To answer the question, should [compulsory] face mask wearing [in some settings] be recommended? Well it’s unfortunate that “there are currently no studies that measure the impact of any kind of mask on the amount of infectious SARS-CoV-2 particles from human actions” but we wouldn’t want to let that fact lead us to an answer which would rightly attract mockery, would we? The precautionary principle is being invoked after consideration of all the relevant evidence.

      • Luca:

        The case reports make it irrefutable. When one person infects 30-50 other people at a choir practice without touching a single one them, it’s hard – impossible – to believe that every person ingested non-aerosol particles from a sneeze or a cough. Similar situations are documented in many CDC case reports.

        In my world, I select the most likely phenomenon to cause the apparent result. In WHO’s world, apparently, there’s a preferred phenomenon – coughing and sneezing – that must be overturned by evidence in which every molecule must be traced into the body of the infected person and then demonstrated to cause subsequent infection.

        Besides the lack of recognition of the actual mode of transmission, WHO and other experts aren’t doing anything to tamp down inane beliefs about transmission. Just recently, someone came to my house and took of their shoes, pointing out that COVID could come in on the soles of shoes!!! OMG!!! INVINCIBLE COVID!! EASILY REMOVED BY SOAP AND WATER BUT MYSTERIOUSLY SURVIVES HOURS OF BEING SCRAPED AGAINST PAVEMENT TO BRING DEATH AND DESTRUCTION WHEN FAMILY LICKS FLOOR TOGETHER!!!

      • “at this point what’s needed is hard, experimental evidence.”

        How about some evidence that it’s transmitted by coughing and sneezing? Go find one case report where it was demonstrated that coughing and sneezing caused transmission. Find just one.

        REALITY: there’s less evidence for coughing and sneezing as the primary form of transmission than there is for aerosols.

        • > a variety of organizations were flogging all kinds of nutcake analyses and making wild and unsupported predictions from them to guide policy, all of which had pretty much zero positive impact despite the hours of chin-rubbing they generated, and created a major distraction from the relevant information;

          I don’t think this is coherent. It just seems like there’s way too much certainty in what you’re saying. So all modeling efforts led to zero positive impact? But somehow there was relevant information that was not subject to modeling assumptions?

          > Go find one case report where it was demonstrated that coughing and sneezing caused transmission.

          No way your error bars can be this small on the statements you’re making!

          The aerosol theory includes coughing and sneezing, right? Like coughing and sneezing can produce small particles as as well as large particles?

          The difference in the droplets and aerosols argument is that the aerosols are tiny little things that stay suspended in the air for a long time and droplets just kinda sink?

          Isn’t the justification for masking attached to the droplet theory? The aerosol theory tends towards masks are useless because it’s background levels of the virus that are floating in the air (and so it doesn’t matter if you have a mask in a badly ventilated room).

          I still like your big ventilated room vs. small unventilated room transmission story, but I don’t think that requires any laser precision dichotomy in particle size theories.

          > But the fact is that even much of that could have been prevented if CDC and many other “experts” hadn’t been so stuck on a pre-determined idea about the transmission mechanism and completely oblivious to the accumulating evidence of asymptomatic transmission.

          Gaah, gimme a case study where an “expert” caused transmission. Do the experts only need to be maintained at an ambient concentration in the room to be infectious? Or do the experts need to land in the noses/mouths of the to-be-infected?

          I don’t like caping for these big, random organizations cuz I don’t really know how they work or what they do (like what the hell is this: https://www.thedailybeast.com/dr-anthony-fauci-is-an-instyle-cover-starjust-as-he-should-be), but the covid response at least from my corner of the world has been incredibly fractured and confusing, and so an analysis that goes back and lumps all the blame on the WHO/CDC/”experts” seems incomplete.

        • “So all modeling efforts led to zero positive impact? ”

          Absolutely without question.

          There has been no positive public health impact from modelling. If you have even a qualitative suggestion to the contrary, by all means share. I don’t demand absolute proof or even numbers. Just support the case for your favorite model with a rational argument. There may be a minor positive scientific impact: there’s an increasing realization that stupid models are stupid.

          “No way your error bars can be this small on the statements you’re making!”

          Exactly the point. There’s no evidence that supports the “large droplet” theory over any other theory. It’s just presumed to be so.

          “Isn’t the justification for masking attached to the droplet theory? ”

          Not at all. The mask can delay infection in an aerosol environment. It’s not known how long. But in most case reports exposure time is an hour or more.

          “Do the experts only need to be maintained at an ambient concentration in the room to be infectious? ”

          Experts are infectious at any concentration. Everyone wants to be the one calling the shots! Me too! :)

          “an analysis that goes back and lumps all the blame on the WHO/CDC/”experts” ”

          Gov officials have done a terrible job. But I think it’s appropriate that the people who called the shots take responsibility for the shots they called. Scientists and leading health orgs are supposed to evaluate the evidence, not spout cliches.

  3. I’d like to get the discussion back to data, and away from testing and policy. Does anybody know why we had such lousy CDC data collection in March and why it hasn’t improved? And don’t cite Trump: this is a problem deep within the CDC that has to be longstanding, so his only sin is in not doing massive demotion of the civil service leadership there.

    For example, I looked up this weekly death data recently:

    Date covid deaths
    4/18 17K
    6/20 3
    6/27 1

    https://www.cdc.gov/nchs/nvss/vsrr/COVID19/index.htm

    It looks like deaths are down over 90% from 17K to 3 to 1, right? But then I looked at the notes under the table and found that they report deaths as the files come in from the states and it takes up to 8 weeks for the weekly data to arrive. So, actually, this CDC webpage is useless for anything about deaths in the past 2 months. This has two bad parts: (a) they don’t collect the most basic, easy, relevant, data on covid-19 in time to be useful, and (b) they give us some other numbers that are irrelevant but will fool people who don’t read the fine print into thinking they have the real numbers.

    This has been a bit long, but I really would like to hear if anyone knows what’s going on down there in Atlanta. How can data collection be so bad? Washington DC economics statistics bureaucrats look like angels by comparison, so it isn’t something about the nature of government generally.

    (https://www.rasmusen.org/blog1/covid-19/)

    • One component of all this is that huge quantities of medical data are still sent on REAMS of Fax paper due to a complete fiction that somehow this is a “safe” mode of transporting sensitive information.

      A law making it illegal to send medical data by fax, and requiring all medical data to be transported by electronic means over a medium with an approved level of encryption would be a simple fix for this, and virtually overnight you’d have people uploading data on websites over TLS.

    • CDC is a bureaucracy that hasn’t been seriously tested in a publicly visible way in a long, long time. The US hasn’t had really bad infectious disease outbreaks since AIDS was new; even 2009-10 swine flu turned out to be not that scary. Given that, functioning fairly poorly is not surprising.

      And the CDC is just not set up (anymore, at least) for quick response. For the things they’ve normally been doing, a 2 month delay is unimportant. The US hasn’t had serious fast-moving infectious disease outbreaks for *ages* before COVID, so they’ve put a lot of their attention into lifestyle diseases like obesity, smoking-related things, etc. (where time isn’t critical) and things like Ebola (where their success or failure is not easily determinable by the public; did Ebola not spread in the US because CDC succeeded, or because the conditions were just not there for it to do so in the first place?).

  4. Prof. Gelman,
    Wondering what your views are on the lack of accurate data on Covid 19 fatalities? Toronto public health authorities tweeted a few weeks ago that patients with Covid-19 who passed away were added to number of Covid-19 deaths, even if cause of death was different.

    Similar argument of data inflation made here: https://www.spectator.co.uk/article/the-way-covid-deaths-are-being-counted-is-a-national-scandal

    If the data itself is of such bad quality (with false positives in diagnosis + fatality numbers), doesn’t it put the debates using statistics on shaky ground?

    GIGO.

    • All data has quality issues, the existence of a few people who are classified as COVID deaths but were actually hit by a bus or choked to death by a police officer really isn’t indicator of anything. There are WAY WAY more people who died from COVID but were not diagnosed early on due to lack of tests for example.

      The real question is what are the sizes of the errors an the biases, and right now everything suggests massive UNDER diagnosis and that probably includes under diagnosis of deaths.

      • Certainly COVID deaths were well undercounted early in the pandemic in the US, as seen from CDC excess death curves (lots of excess deaths unexplained by COVID). So total COVID deaths are probably undercounted. But the CDC excess death curves seem to show this undercounting stopping sometime in May (after that there are not excess deaths unexplained by COVID). But CDC data lags enough that who knows, really?

        Differences in counting of deaths probably do mess up comparisons between nations. It’s entirely possible that France has actually had fewer genuine COVID deaths per capita than the US, for example, though the JHU numbers show it slightly higher.

        • Certainly COVID deaths were well undercounted early in the pandemic in the US, as seen from CDC excess death curves (lots of excess deaths unexplained by COVID

          Here is cumulative deaths by week for 2014-2020 from cdc: https://pastebin.com/aPLMv2dY
          And here is cumulative covid deaths by day for 2020 from covidtracking: https://pastebin.com/G6ni2VUm

          It generally increases year on year, except 2018 had more deaths than 2019. Compared to 2018, I see 23,341 excess deaths by week 15 (ending april 11th 2020). By April 11th covidtracking reports 23,130 covid deaths. I picked that date because it was the peak week for 2020 deaths. For week 16 it was 12k more for all cause, week 17 it was 18k, week 18 it was 16k. Etc.

          So up to the middle of may, looks like only 10-20k deaths can be missing. Maybe someone wants to do a better analysis than I felt like at the moment.

        • Yeah by “lots” I didn’t mean necessarily more than 10-20k. Just that it was enough to be easily visible on the excess deaths curve: https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm

          The worst week for deaths is week ending April 11, with 8.5-12.4% excess deaths unexplained by reported COVID deaths, 36.5-41.4% excess deaths overall. So even in the worst week about 3/4 of the excess deaths are explained by COVID, much more most weeks. So yeah an undercount of more than 20k doesn’t look that plausible. We’re probably really at say 150k deaths now.

      • So the assumption you’re making is:

        Under-diagnosis due to lack of testing > over-diagnosis due to the conflation of symptoms with cause of death?

        • I think about it like this there’s:

          actual_covid_deaths + error1 + error2 + error3 + …

          if we order error1,2,3 in decreasing order of size, then the issues like people dying of being hit by a bus while coughing is way way down the list… The large errors are under-diagnosis due to lack of testing (current estimates are that only around 1/10 people who have the disease are being diagnosed), and then issues like what “confused” talked about below.

      • Although, on further thought, excess deaths may not be a true measure of undercounting vs. overcounting, and in some cases there may not be an unambiguous right answer.

        Deaths can have multiple causes, and there might be genuinely-COVID deaths that are not “excess” if COVID is, say, displacing other causes of pneumonia in the “ecosystem” of pathogens. Even outside flu season, at least several percent of all deaths are pneumonia, and if COVID is more contagious than the other causes (at least due to lack of vaccines for it, if not inherently) there might be some ‘displacement’.

        Also, there may be heart attacks caused by COVID, and heart attacks caused by other problems that lead to death when they wouldn’t normally due to fear of COVID keeping people away from ERs.

        • Indeed, it’s all very complicated, and the existence of a number of people who obviously died of being hit by a bus while coughing is not really even on the radar in terms of the size of errors.

        • Yeah. Among relatively young relatively healthy people, the background chance of dying in a particular month is pretty low, so there won’t be many people who happen to incidentally die of unrelated causes while having COVID. There will be some, but not enough to materially affect the overall statistics.

          The more complex questions are things like people who were already pretty likely to die in the next month, get COVID, and die from an immediate cause other than obvious pneumonia. Sure COVID can cause deaths other than pneumonia deaths, but if that person was already likely to die of cause X, attribution gets trickier.

          I don’t think this is going to be significant enough to affect the overall picture, or balance out the undercounting earlier in the pandemic — but it may distort the shape of the curve (IE, if deaths were significantly undercounted in March and slightly overcounted in June, the decline from March to June will look less dramatic than it was in reality).

      • That’s not all-cause, that’s influenza like illnesses and pneumonia.

        COVID seems to be pretty clearly at this point a systemic disease, virus damage is found in the brain, kidneys, liver, heart, digestive tract, etc. The Pneumonia deaths is likely to somewhat undercount deaths actually (and if you look at the data, it’s a HUGE spike so if that’s an undercount… geez). In particular, heart attack and strokes seem to be caused by COVID and they won’t be in there. There may also be longer term issues like kidney failure in people who had marginal kidney health before COVID etc.

        • That’s not all-cause, that’s influenza like illnesses and pneumonia.

          All cause is there if you download the data.

          COVID seems to be pretty clearly at this point a systemic disease, virus damage is found in the brain, kidneys, liver, heart, digestive tract, etc.

          All the symptoms are also symptoms of high altitude sickness, and many strange ones (loss of taste/smell, clotting) were said to start resolving upon HBOT therapy. So I suspect they are all downstream from the same thing that makes it mimic high altitude sickness.

        • According to this Cornell page (it links to the study – hopefully reliable) https://news.weill.cornell.edu/news/2020/07/strokes-occur-more-frequently-in-patients-with-covid-19-compared-to-flu-but-overall stroke risk from COVID is 7x that from flu. CDC does say flu raises risk of heart attack and stroke.

          Since COVID IFR is maybe 0.5%-1.1% or something, and flu is 0.1% or less, a 7x factor in stroke risk seems to be pretty compatible with the overall risk increase from COVID. IE – do we know that COVID is “more” systemic than flu, or do we just see more systemic effects/complications because COVID is deadlier overall?

  5. Just a general comment that I really appreciate this blog. Many interesting and relevant topics.

    Of course, the blog owner, and the other contributors, but also the commenters.

    High percentage of quality comments, people making an effort to base ideas on data, rather than tribal values.

    A blog oasis. Thanks!

Leave a Reply to Anoneuoid Cancel reply

Your email address will not be published. Required fields are marked *