Skip to content

“Estimating the effects of non-pharmaceutical interventions on COVID-19 in Europe”

Seth Flaxman writes:

Our work on non-pharmaceutical interventions in 11 European countries (originally Imperial report 13) is now published in Nature, Estimating the effects of non-pharmaceutical interventions on COVID-19 in Europe.

Of note for your readers:

1) Nature has an open peer review process, so you can see the (pre-publication) peer review here.

2) Between submission on 30 March and publication we updated our model a few times and of course regularly refit it as new data became available. In order to understand what impact new data had, we retrospectively refit the final model to each new day of data, and created these animations to show how the posterior contracts.

3) Stan source code for this and subsequent reports is all available—thanks to the community for improvements and speedups!

Subsequent reports and a paper:

USA: and website: and open review:



“Have deaths from COVID-19 in Europe plateaued due to herd immunity?” in the Lancet:

4) Stay tuned for an R package for our model, based on RStanArm. If you’re interested in helping beta test it, please get in touch.

We discussed some version of this model last month.


  1. James Annan says:

    In principle I like this work, I think it’s a really neat idea. However I have difficulty avoiding the conclusion that it’s substantially overfitted. The 95% credible intervals appear to exclude far too much of the data, and the obvious inadequacy of the model doesn’t seem to be accounted for in the estimation. Attributing the vast bulk of the effectiveness to one intervention and basically nothing to any of the others seems a bit optimistic to me. They were implemented in a different fashion across countries, all within a relatively close window, and yet the estimates of their effectiveness are converging to incredibly sharp estimates many weeks later.

    I realise what I’m saying isn’t necessarily compelling evidence but it certainly looks just like what I’d expect to see in an overfitted model.

    • Daniel H. says:

      I agree with this criticism. I did look at the paper a few days ago and while I’m not qualified at all to weigh in on this, here are a few points I did notice:
      Things I liked: Data and code are all open. I cannot stress how much this helped – even for me as an outsider in a leisure-time discussion of the results with friends.
      After looking at your code, I found that case mortality rates appear to be preset to hard values (mostly 1-1.2% mortality for various countries). This makes sense since these numbers are almost unidentifiable as of today and there’s a short mention of comparing them with antibody studies, but I’m really really concerned that I had to spend 15 minutes searching in your source code to find these numbers – while they are a central assumption in the whole model and directly scale the results regarding infection rate (at least if my understanding is correct – infections are basically estimated by dividing deaths by infection rate and accounting for time offset).
      Also, the model assumes a set of fixed interventions without any other varying effects, while we already know that there are numerous, e.g. problem awareness in the population (explaining e.g. why Sweden is having a roughly constant number of new cases over time without lockdown or school closure) or variation of the number of tests over time (which probably explains why case numbers in Sweden are rising in the last few weeks; I would not expect death numbers to follow).

      In summary, I guess the estimates on the lives safed by lockdown are possibly off by an order of magnitude, but the model and paper has still some worth as providing a reasonable upper bound on the numbers. But I’m missing all this in the paper itself (this should be really prominent right in the summary!) and I’d be interested to see what others think.

      General warning: I’m not qualified in any way other than reading this blog on reviewing the paper – so pleae read my comment with the needed scepticism.

  2. Adrian Keister says:

    Isn’t voluntary staying-at-home an extremely significant confounding variable? Your own post here:

    essentially showed that: in the US, at least, the people were ahead of the government practically every step of the way; in that case, it couldn’t have been the lockdowns that did anything, because they didn’t change anyone’s behavior. Instead, if anything, it was people voluntarily staying at home. Was it like that in Europe as well?

    To me, the paper can’t prove what it says it proved. If the paper were to be rewritten to say that voluntary staying-at-home has been effective at preventing COVID deaths, it might have something. Its conclusions:

    “Our results show that major non-pharmaceutical interventions and lockdown in particular have had a large effect on reducing transmission. Continued intervention should be considered to keep transmission of SARS-CoV-2 under control.”

    are unwarranted.

    The analysis of Sweden was rather disappointing, as well. All the paper said was that, due to dissimilarities with other countries, there was more uncertainty in the results. Whoa: don’t go out on a limb, there. The graphs on pdf page 11 do not appear to be per capita, which makes comparing one country to another rather difficult.

    It’s been over 3 weeks since George Floyd died, and the large groups of rioters and protesters have certainly been disregarding any lockdowns. 3 weeks should have been enough time see if there was any rise in COVID cases. What we see instead is a negative second derivative before and after: new cases and deaths are both slowing down, and there’s no apparent response to the riots and protests.

    In the acknowledgements, it shows that the authors received support from the Bill and Melinda Gates Foundation – disappointing that yet another study gets money from a far leftist organization like that one. While it’s good, and even ethical, that they acknowledge that funding source, it’s certainly not un-heard of for such money to color the results. To be clear: far leftists want control and power. If the far leftists give money to some scientific endeavor, there is a decided risk that the conclusions will be determined ahead of time: support the far left agenda. Otherwise, the money might not have arrived in the first place.

    In summary: no. I’m not inclined to give this paper much, if any, weight.

    Adrian Keister

    • jim says:

      Whoa, our old friend, the “S” word!

      “Our results show…”

      Absolute certainty!

    • oh come on says:

      I don’t want to turn this into a food fight, but it is absolutely comical that you think a billionaire’s bland center-left think tank that mostly promotes technocratic (some would say neoliberal) solutions is “far left.”

    • Dean Eckles says:

      Pre-order mobility reductions and individual choices as to level of compliance with orders do seem to play a big role in observed mobility reductions. We attribute a bunch of this variation to social influence

      “Bill and Melinda Gates Foundation – disappointing that yet another study gets money from a far leftist organization like that one”

    • Adrian Keister says:

      Concerning the Gates Foundation being far left: I could well be wrong about the Gates Foundation being far left. It’s definitely leftist, and the criticism I leveled at the paper would still hold even if the Gates Foundation is leftist but not far left. This is really quite a minor side-point, especially compared with the glaring confounding variable issue which I regard as fatal.

      • dhogaza says:

        Adrian Keister

        Gratuitously inserting your political views as you did makes it difficult to take you seriously. Sorry.

        • Adrian Keister says:

          The issue is already political, simply by virtue of talking about lockdowns – a government intervention. I find it much more honest and direct to come out and say what your bias is; mine, of course, is heavily conservative. In an inherently political discussion, having biases out in the open is much better. So that, in your case, for example, it’s much easier, once you find out that I’m conservative, to not “take me seriously”.

          Everyone is biased. That’s not necessarily a problem. Not stating what your (general “your”, not dhogaza’s) bias is and claiming to be objective is a problem.

          So I would actually reverse your (dhogaza’s) statement, and say that, in an inherently political discussion, failing to state what your (general “your”, not dhogaza’s) political views are makes it difficult for me to take you (general again) seriously.

    • > What we see instead is a negative second derivative before and after: new cases and deaths are both slowing down, and there’s no apparent response to the riots and protests.

      what universe are you in? Cases have been rising sharply in many states. For example SC, AZ, CA. Most of the people involved in the protests were younger though, so will be more likely to be asymptomatic or have mild symptoms and not seek hospitalization etc.

      • Adrian Keister says:

        Well, I guess I was just looking at MN data (should have mentioned that in my previous comment, so thanks for pointing that out), where the riots and protests have been more heavily concentrated than in most places, though no doubt there are more heavily concentrated riots and protests elsewhere. They didn’t seem to make any difference at all in MN. The case rates are definitely slowing down in MN.

        I should also mention that I, for one, am really not terribly interested in cases. I can’t get worked up over healthy people getting COVID and recovering in a week or two. Deaths I care greatly about. In SC, death rates have definitely not been on the rise, though I will grant case rates have. In AZ, case rates are rising, but it’s unclear whether death rates are rising. The case rate does seem to have steepened a few days after the death of George Floyd, which is suspicious. In CA, the case rates might have taken a slight steepening, but the death rates appear actually to have a slightly negative second derivative.

        Because of no observable difference in death rates, it appears to me that the riots and protests have not had an effect on death rates.

        • The problem is death rates are really quite lagged from infection rates. Early in the pandemic they appeared to be not that lagged, but that’s because case ascertainment was at a much LATER stage in infection in my opinion. Now we are catching cases 5 or 7 days from infection, but back in March we were catching cases 1 or two hours before being put on a ventilator and then dying… which was probably 20 days after infection. (for example in NYC).

          Also, in regions where prevalence was low, even big riots and protests wouldn’t have spread much. On the other hand where prevalence was higher, you could get dramatic spreading. Southern CA is an example.

          Hospitalization in TX and AZ is climbing fast. The US is huge, even state level data is too aggregated in many cases. For example SoCA is a different region than SF Bay and both are different from Sacramento, or Redding or Bakersfield.

          • Martha (Smith) says:

            “The US is huge, even state level data is too aggregated in many cases. For example SoCA is a different region than SF Bay and both are different from Sacramento, or Redding or Bakersfield.”

            Yup. We (the US) still have a long way to go in collecting appropriate and useful data, curating it well, and using it well.

          • Navigator says:

            The increase they have been reporting lately are largely a function of more wide-spread testing. Earlier in April or May, even with available testing sites, there was a lot of confusion as to who/when/why should get tested, so only symptomatic and nervous Nelly types did it.

            Past few weeks, testing has been more widespread, so the apparent increase is only showing what’s been there all along.

            I agree that in March, the infection detection was done too late on per person basis (just before being hospitalized), but I don’t think we’ll see huge spikes in deaths, because of behavior modification that has happened since.

            Most people are now aware of distancing/masks/etc., all public places are sanitized to death, hours reduced, crowds controlled. None of that was present in Feb./March, which resulted in those huge death spikes then.

            But yes, it will go up and down for a while until the virus gets bored and leaves us.

    • Joshua says:

      Adrian –

      > it couldn’t have been the lockdowns that did anything, because they didn’t change anyone’s behavior. Instead, if anything, it was people voluntarily staying at home.

      I’m guessing you don’t work for an hourly wage? Aren’t getting unemployment after losing your job? Don’t have kids?

      What would people have done over time without government mandated shelter orders in place? W/o them, would people have been able to stay home to take care of kids? Would they have had to choose between staying safe and losing their job? They could have chosen to stay home and get fired and collect no unemployment, or go to work and bring home infection to grandma? No curbside service for vulnerable people? Schools not closed so kids have to go, teachers above 60 have to teach. Does anyone get stimulus checks? How many people would have been evicted?

      I think the shelter in place orders enabled a lot of things to happen.

      • Adrian Keister says:

        Joshua –

        Your questions about whether I work for an hourly wage, am getting unemployment, or having kids are quite irrelevant.

        Well, the lockdowns caused the job losses and business closures in the first place; it’s a bit weird to say that “A caused these wonderful things, C, to happen!” when A first caused some extremely bad things B to happen, for which the C are partly a solution.

        • Joshua says:

          Adrian –

          I was speculating that a lack of relevant life experience might help explain the oversights in your argument.

          > Well, the lockdowns caused the job losses and business closures in the first place;…

          I think you missed my point. Many of those jobs may well have been lost anyway, becsause of the economic impact of a raging virus. And with no IOS orders ploywrs would have had to fire people without getting any of the loans fro the PPP. Or many people would have been forced to be exposed risk they didn’t want to take because if they didn’t go into work they would get fired and couldn’t collect unemployment. Or many people would have been forced to send their kids to school even though they didn’t want to be sue they couldn’t stay home to be with their kids because if they stayed home they would get fired and wouldn’t be able to collect unemployment. I suspect thatanh owolw wouls have chosen to be safe rather than expose themselves and their families to thar risk. And thus, the SIP orders “did something” because it enabled them to reduce their risk, not lose their jobs, and collect unemployment.

          Its too simplistic to just say that the SIP orders cost people their jobs. The virus played a role also. For people who are hourly employees and couldn’t just work from home, no SIP orders would have forced them to make terrible decisions, and cost some of them their jobs with no backup thst they got as a results of the SIP orders.

          • Adrian Keister says:

            Joshua –

            > I was speculating that a lack of relevant life experience might help explain the oversights in your argument.

            Well, I would recommend against that sort of speculation, as it reflects no credit on anyone. No doubt there are gaping holes in my argumentation, but poisoning-the-well fallacies of the sort you’re leveling against me will not “explain” them. Your statement also assumes there are oversights in my argument.

            If you compare Sweden’s unemployment with that of the US, you find that, while there certainly does appear to be higher unemployment, possibly due to COVID, it’s not nearly as high (Sweden in April, e.g., was at 8.2, compared with the US at 14.7; then Sweden in May was at 9, compared with the US at 13.3. Source:; I have no idea if that’s a reliable source, but it was convenient.) So, while you’re quite right that it appears the virus is causing significant unemployment, the lockdowns appear to be causing yet more significant unemployment. Moreover, with the lockdowns in place, options are more limited for private individuals to solve their problems.

            And this is really one of the main political points I try to make: it simply is not the government’s job to solve the virus problem. The American people, for example, were ahead of the government every step of the way (my main point in my first comment, and by far the most important; I note that no one appears to have interacted with it at all, but are picking away at much less important matters). The lockdowns did not change people’s behavior in staying at home, and could therefore have had no causal impact on preventing the spread of the virus. It’s an elephantine confounder that, so far as I’m concerned, completely invalidates the paper.

            It is not the government’s job to bail people out. The Keynesian economics principle of injecting non-existent money into the economy is not going to solve any problems, but it will create them.

    • Jim says:

      Yes, I agree with most ppl, Gates foundation isn’t “leftist” at all.

    • Joseph Candelora says:

      I’m not sure what your purpose is for hair-splitting over lockdown.

      If everyone quarantines at home and wears a may because they choose to, or everyone quarantines at home and wears a mask because the government mandates it, what difference does it make for analysis like this?

      • I think there are two different questions:

        1) the bio-physical question of how effective is it for people to stay home in terms of reducing the number of people who get infected per unit time

        2) the sociological question of how effective is it for governments to *mandate* that people stay at home in order to receive whatever benefits there are in (1)

        People aren’t reading these studies with the same questions in mind when they are asking about effectiveness… some are thinking (1) and others are thinking (2)

      • Adrian Keister says:

        I agree with Daniel, and I would add that there is an absolutely gigantic difference between the two, and it is not hair-splitting at all. The difference is the economy, as well as political freedoms. Many people, such as myself, value a vibrant economy unfettered by too much governmental intervention (some you certainly need, to prevent theft). Many people, such as myself, also highly value our freedom.

        The argument of the paper is clearly in favor of government-mandated lockdowns, which is a wildly different thing from people voluntarily staying at home. And I am saying that, in terms of virus spread, the paper’s argument that it is the government-mandated lockdowns that are effective in preventing the spread of the disease (and therefore the government-mandated lockdowns should continue) falls flat on its face for the reasons I have given above.

        • Joseph Candelora says:

          “The argument of the paper is clearly in favor of government-mandated lockdowns, which is a wildly different thing from people voluntarily staying at home. ”

          You’re going to have to support this “wild difference” between the two things. You said yourself that they’re effectively equivalent — “the people were ahead of the government practically every step of the way; in that case, it couldn’t have been the lockdowns that did anything”.

          You’re splitting hairs of no consequence here. When NY State announced closure of non-essential businesses on March 20, my company put everyone in all states across the country on work-from-home status. McDonald’s had already closed all dining rooms in company-owned stores on March 16. These are the actions that dropped the transmission rate. They’re also the actions that caused layoffs and the drop in economic activity. How exactly is you’re arguing that the “difference is the economy”?

          There’s a much stronger argument that individuals’ and businesses’ overreaction to conditions in limited areas caused unnecessary additional damage to the economy nationwide vs limiting Covid reactions to just those mandated by the state.

          • Mendel says:

            My personal experience is that people are ahead of lockdowns because lockdowns are announced in advance. Cancelling large events and announcing school closures for the coming week communicated to everyone that this is serious, and we must act now; consequently, at a time when restaurants and cinemas were still open, they were already empty (I was there); people took social distancing and hygiene seriously in part because the government was.

            > “individuals’ and businesses’ overreaction to conditions in limited areas caused unnecessary additional damage to the economy nationwide”

            I believe it is difficult to differentiate an “overreaction” and “unnecessary damage” from “pitching in to do their part” and “unavoidable damage”: if you believe in freedom, you have to also believe in responsibility.
            From the outside, the predominant problem of the US seems to be *under*reaction causing unnecessary damage: in “the hammer and the dance”, the hammer hasn’t come down hard enough, and that prolongs the period of high prevalence of infection. Resisting the cure because “liberties” means you stay sick longer.

            If you’ve found an easy way to distinguish overreaction from underreaction, let me know.
            The only way I know to tailor the reaction to the prevailing conditions is monitoring and using feedback based on that. Meanwhile Trump is on the news with “if we don’t test, we have no cases”. Go figure.

            • confused says:

              >>My personal experience is that people are ahead of lockdowns because lockdowns are announced in advance

              In my state (TX), people, businesses, and local governments (school districts, cities, and counties) acted way before any announcements of stay-at-home orders. School districts in the Dallas/Fort Worth area were closing as early as March 12-13, which then led employers to act, etc. The large cities/urban counties also put in measures before the statewide government.

              Another factor is that many large businesses operate in multiple states and nations, and some announced company-wide telework policies, etc. even in states that were not talking about lockdowns.

              But my employer is Texas-only, and we went to telework over two weeks before the statewide order.

              Of course, TX is a large state and has a huge urban/rural density variation compared to states like MA, NJ, CT on the one hand or WY, MT on the other. Arguably it made sense to impose statewide measures significantly later than measures specific to huge cities that are major international ports-of-entry (Dallas and Houston).

              • confused says:

                I think some of the issues* Texas is having now is because we cut down activities very early on in the local spread (first reported COVID death in TX was March 17, after school closures were already in place in the larger cities and many employers had moved to telework).

                So the early spread was pretty limited, and it seemed like COVID was not really an issue here, and we started to run into economic and social/political issues (TX is very individualistic culturally and very pro-business politically) maintaining the stronger measures, and we reopened.

                If we’d waited a few more weeks to do the stay-at-home order, things might have turned out fairly differently.

                *In the sense that I think that COVID is actually increasing in Texas fairly quickly (as shown by hospitalizations), but I also think that better testing than what we had earlier on makes it look even worse. The higher deaths from the end of April to mid-May suggest that early-mid April had a ton of un-noticed infections (“a ton” compared to the number of confirmed cases, not compared to say New York).

  3. Carlos Ungil says:

    I’ve not really looked at it, but I remember reading “Report 13” and my conclusion was that modeling at the country level was a huge limitation. The orginal prevalence estimates and credible intervals were in fact quite incredible for Spain and Italy.

    As of 28th March, 15% [4% 41%] for Spain and 10% [3% 26%] for Italy. Those numbers were hard to reconcile with reality: getting those prevalences at the country level assuming that regional differences were in line with the observed variation required a large majority of the population in Lombardy or Madrid to be infected.

    (The estimates in the revised paper are more reasonable: 5.5% [4.5% 7%] for Spain and 4.5% [3.5% 6%] for Italy.)

  4. Daniel H. says:

    I’ll borrow this post for a comment unrelated to this specific paper but very much related to covid-19-modelling:
    I recently came across the causel models by the group around Karl Friston, see
    for the interview that sparked my interest and
    for publications and model code.

    Based on my first impression, their approach appears to be something along:
    1) use timeseries modelling (kalman filtering? generalized filtering? I’m not really sure on how to call that technique) to fit various models with hidden / latent variables to the epidemic data (test numbers, cases, deaths, interventions over time for various countries)
    2) rate models based on how accurate they predict the further pandemic development (bayesian model evidence, if I get it right this basically favors the simplest model that gives accurate predictions. As more data accumulates, this will shift the preference from very simple, low-parameter models to more detailed models)
    3) predict the near-future development by a weighted average of models with weights according to model evidence (I don’t really know if this is what they are doing, it’s just what I would do at this point; from an epistemic point I wouldn’t want to put all predictive weight on just one model but instead get several opinions, but listen a lot more to the opinion with the best previous track record)

    Can someone comment on this approach? Did I get it right or am I missing something important? Is this a standard procedure or totally unusual?
    And a quick check on model evidence did bring up a lot of references on bayes factors (which I’ve seen on this blog mainly as a “bayesion flavor of p-values” we’d like to avoid as well), so are their serious drawbacks or limitations to the route outlined above?
    Cheers, Daniel

    • Ben Bolker says:

      I keep meaning to come back and post something about Friston, but I keep hesitating because of the time and energy it would take to do it properly. Lazily: I spent about an hour reading Friston’s paper (for a lab meeting discussion) and was underwhelmed. There are some nice technical aspects, but the bottom line is that Friston **strongly** underestimates the uncertainty in his results: a couple of quick examples are that he (1) assumes that variation in case reports is strictly binomial (e.g. no underlying variability that might make it beta-binomial or overdispersed in some other way); (2) I believe that a lot of the results quoted depend on the posterior distributions all being multivariate Normal; (3) he assigns particular functional forms for the effects of susceptible depletion (“dark matter”) and the effects of behavioural reponses and concludes on that basis that susceptible depletion rather than behavioural response is driving the epidemic.

      A physicist’s commentary on this work is here: . It concludes: “… I would argue that the “dark matter” in Friston’s model extends well beyond the immunological aspects. It would need many more data points and a much more extensive description of the causes affecting them to forecast the pandemic in any accurate way.”

      On the good side, the code is available (I haven’t tried it out):

      Friston has been involved in conversations about sample size and statistical methodology: a critique is here:

  5. Guillermo D. says:

    Today came out this post from Nic Lewis that I think you may find interesting to read (to say the least).
    He argues against the Flaxman et al. methods, and since the third paragraph you may find a lot of funny words, but I’m sure it can bring some discussion here:

  6. Joshua says:

    A climate “skeptic” weighs in to tell us that the Nature article got it wrong. I’d be curious if anyone from this crowd (given that y’all can follow the technical arguments) would care to commentn



    First and foremost, the failure of Flaxman et al.’s model to consider other possible causes apart from NPI of the large reductions in COVID-19 transmission that have occurred makes it conclusions as to the overall effect of NPI unscientific and unsupportable. That is because the model is bound to find that NPI together account for the entire reduction in transmission that has evidently occurred.


  7. Anoneuoid says:

    Covid in March 2019?

    Most COVID-19 cases show mild influenza-like symptoms (14) and it has been
    suggested that some uncharacterized influenza cases may have masked COVID-19
    cases in the 2019-2020 season (11). This possibility prompted us to analyze some
    archival WWTP samples from January 2018 to December 2019 (Figure 2). All samples
    came out to be negative for the presence of SARS-CoV-2 genomes with the exception
    of March 12, 2019, in which both IP2 and IP4 target assays were positive. This striking
    finding indicates circulation of the virus in Barcelona long before the report of any
    COVID-19 case worldwide. Barcelona is a business and commerce hub, as well as a
    popular venue for massive events, gathering visitors from many parts of the world. It is
    nevertheless likely that similar situations may have occurred in several other parts of the
    world, with circulation of unnoticed COVID-19 cases in the community.

    • Almost certainly this is false positive from contamination. This issue is well known by ecologists who sometimes use PCR to look for the presence of invasive species, but are winding up with false positives from say 5% of PCR tests just due to background contamination in the lab or lab technician accidental contamination or whatever.

    • Joshua says:

      > Covid in March 2019?

      Wait, but I’ve been reading in the rightwing media and on the IDW that there’s clear evidence that covid 19 comes from a lab in Wuhan!

      • Anoneuoid says:

        Pathogens escape from these labs all the time. And these arent independent events (ie most will come from the same labs with lax security):

        The SARS virus has escaped from high-level containment facilities in Beijing multiple times, notes Richard Ebright, a molecular biologist at Rutgers University in Piscataway, New Jersey.

        It is really dumb to have these BSL-4 labs near major population centers. If people want to work on that stuff put it out in the middle of nowhere and have a month quarantine before you rejoin the population. It will make sure careerists stay away from the project too.

      • Anoneuoid says:

        SARS has apparently escaped these labs 6 times that we know about as of 2014:

        SARS has not re-emerged naturally, but there have been six escapes from virology labs: one each in Singapore and Taiwan, and four separate escapes at the same laboratory in Beijing.

        The first was in Singapore in August 2003, in a virology graduate student at the National University of Singapore. He had not worked directly with SARS, but it was present in the laboratory where he worked. He recovered and produced no secondary cases. The World Health Organization formed an expert committee to revise SARS biosafety guidelines.

        The second escape was in Taiwan in December 2003, when a SARS research scientist fell ill on a return flight after attending a medical meeting in Singapore. His 74 contacts in Singapore were quarantined, but again, fortunately, none developed SARS. Investigation revealed the scientist had handled leaking biohazard waste without gloves, a mask, or a gown. Ironically, the WHO expert committee called for augmented biosafety in SARS laboratories the day after this case was reported.

        In April 2004, China reported a case of SARS in a nurse who had cared for a researcher at the Chinese National Institute of Virology (NIV). While ill, the researcher had traveled twice by train from Beijing to Anhui province, where she was nursed by her mother, a physician, who fell ill and died. The nurse in turn infected five third-generation cases, causing no deaths.

        Subsequent investigation uncovered three unrelated laboratory infections in different researchers at the NIV. At least of two primary patients had never worked with live SARS virus. Many shortcomings in biosecurity were found at the NIV, and the specific cause of the outbreak was traced to an inadequately inactivated preparation of SARS virus that was used in general (that is, not biosecure) laboratory areas, including one where the primary cases worked. It had not been tested to confirm its safety after inactivation, as it should have been.

        • Joshua says:

          COVID 19.

          The “made in a lab in Wuhan” theories circulatng in the rightwingosphere and the IDW are that COVID 19 was created in a lab in Wuhan and got out into public more recently than March of ’19. As far as I know the only remaining thing to be worked pout is whether the release was accidental or intentional

          • Anoneuoid says:

            I am saying that, if it did come from a lab, I’d expect it was released (either on accident or purpose) more than once.

            • Joshua says:

              So if covid 19 (as opposed to other viruses) was released more than once, why is the impact so much greater with the more recent release?

              Mind you, I’m highly suspicious, as I always am with conspiracy theories – esspecially if they’ve gained oxygen in the rightwingosphere and the IDW.

              • Anoneuoid says:

                So if covid 19 (as opposed to other viruses) was released more than once, why is the impact so much greater with the more recent release?

                Most people dont end up being superspreaders so the spread after these types of events is usually self-limiting.

                Mind you, I’m highly suspicious, as I always am with conspiracy theories – esspecially if they’ve gained oxygen in the rightwingosphere and the IDW.

                It is literally something that happens all the time. Many people have been saying it is almost guaranteed eventually a pandemic comes out of one of these labs for years.

              • Anoneuoid says:

                Among the 77 patients, 66 did not transmit to others, and 7 transmitted to 8 others and were designated as associated with superspreading. The pattern of transmission is shown in Figure 3.


              • Anoneuoid says:

                Brackets ruined the quote:

                Among the 77 patients, 66 did not transmit to others, and 7 transmitted to ≤3 contacts. In contrast, four persons (patients A, D, H, and I) transmitted to ≥8 others and were designated as associated with superspreading. The pattern of transmission is shown in Figure 3.

              • Joshua says:

                Only somewhat related – but I thought interesting.

                Preprint – to be ingested with a grain of salt:


                Full genome sequences are increasingly used to track the geographic spread and transmission dynamics of viral pathogens. Here, with a focus on Israel, we sequenced 212 SARS-CoV-2 sequences and use them to perform a comprehensive analysis to trace the origins and spread of the virus. A phylogenetic analysis including thousands of globally sampled sequences allowed us to infer multiple independent introductions into Israel, followed by local transmission. Returning travelers from the U.S. contributed dramatically more to viral spread relative to their proportion in incoming infected travelers. Using phylodynamic analysis, we estimated that the basic reproduction number of the virus was initially around ~2.0-2.6, dropping by two-thirds following the implementation of social distancing measures. A comparison between reported and model-estimated case numbers indicated high levels of transmission heterogeneity in SARS-CoV-2 spread, with between 1-10% of infected individuals resulting in 80% of secondary infections. Overall, our findings underscore the ability of this virus to efficiently transmit between and within countries, as well as demonstrate the effectiveness of social distancing measures for reducing its spread.



              • Anoneuoid says:

                Yep, looks very similar to what was reported for SARS. So 80%+ people do not infect anyone else while ~1-5% infect dozens.

              • confused says:

                Isn’t there also evidence that SARS-COV-2 has not been genetically engineered? (I’m not an expert on this, but I’ve heard that the gene sequence doesn’t look altered.)

                I am *really* skeptical of the lab theory on this. There are tons of animal coronaviruses out there, and many – probably most – human infectious diseases are ultimately zoonotic in origin. (This seems to be why European diseases devastated the Americas but American diseases didn’t devastate Europe — there were far more European infectious diseases since Europeans had more domestic mammals and had lived in close association with them for far longer. The only domestic mammal in most of the Americas was the dog; llamas/alpacas were limited to a fairly local area of South America.)

                The escape of an existing, known disease-causing virus from a lab is quite different from the creation of a new one.

              • Anoneuoid says:

                Im just saying if it came from a lab, it was probably released more than once. And it really isnt hard to make a virus more pathogenic. Look how the mouse adapted strain of SARS was created.


            • Anonymous says:

              I’m not talking about its being particularly difficult – gain-of-function experiments are a known thing – but I really, really doubt that happened here.

              Is there any evidence at all for a lab connection other than that the pandemic started in Wuhan, which has a virology lab? (And I think a lab is there specifically because that area is known for bat coronaviruses…)

              Also, do we even know that it *did* start in Wuhan? The outbreak there was the first *medically detected* but that doesn’t mean the virus jumped from animals to humans there. It could have arrived in Wuhan from a more rural area through someone who had no or mild symptoms and didn’t recognize anything different was happening.

              • Anoneuoid says:

                New paper from this group:

                In contrast, the majority
                of the non-synonymous substitutions in SARS-CoV-2 S are distributed across the gene at low
                frequency and have not been reported to confer adaptive benefit (Figure 4). Yet, the SARS-CoV-
                2 S has been demonstrated to bind more strongly to human ACE2 and has a superior plasma
                membrane fusion capacity compared to the SARS-CoV S (32,33). The only site of notable entropy
                in the SARS-CoV-2 S, D614G, lies outside of the RBD and is not predicted to impact the structure
                or function of the protein (34). Its prevalence in international COVID-19 cases has been attributed
                to the substitution occurring early in the pandemic leading to a founder’s effect. There is no
                evidence of a more virulent strain of SARS-CoV-2 emerging despite passage through more than
                3 million human hosts by the time of this analysis

                The pairwise comparisons of dN and dS, alongside a dearth of signs of emerging adaptive
                mutations, suggest that by the time SARS-CoV-2 was first detected in late 2019, it was already
                well adapted for human transmission to an extent more similar to late epidemic than to early-to-
                mid epidemic SARS-CoV.


                Given that Xiao et al. relied on the 2019 dataset by Liu
                et al. Viruses to generate their genome sequence, it is curious that it was not clear to Xiao et al.
                from the beginning that they were studying the same pangolin CoV.

                Xiao et al. renamed pangolin samples first published by Liu et al. Viruses without citing their study
                as the original article that described these samples, and used the metagenomic data from these
                samples in their analysis.


Leave a Reply