Monday, March 23, 2020

Thirteen Ways In Which The COVID-19 Threat May be Inflated

COVID-19 is a very serious problem that should be taken very seriously by both government and the public, but there are ways in which we may be getting an inflated sense of the threat:


1. In general: Once a dominant narrative is formed (in this case: Covid-19 as huge threat) reporting will be more inclined to cover things that fit that narrative & ignore/dismiss things that seem to contradict it: The evidentiary standards for reporting that fits the narrative will be lower than for reporting that contradicts it.

2. Reporting focuses on *expected* problems rather than currently existing problems: Most hospital capacity reporting is about expected problems.

3. Ordinary events are now portrayed as evidence for the threat, eg:
a) Many hospitals run at near capacity in normal times and will routinely be over capacity but now this problem may be attributed to COVID-19.
b) Rare cases (eg of young people suddenly getting violently ill & dying from COVID-19) that in normal times don't make it into the newspaper (such rare deaths also happen with eg the flu and other viruses) now make the front page & are portrayed as more common than they are.

4. A large increase in the number of cases sounds scary but becomes much less so when this is because of an increase in the number of tests rather than an increase in the number of cases.

5. Death rates can seem very high when everybody who died from other causes but also had COVID-19 is counted as a COVID-19 death (to be sure, this problem is not that widespread, but it is what happened in Italy).

6. If people who died are tested for COVID-19 but not for the flu or the common cold, these deaths may be counted as COVID-19 deaths even though they could have also tested positive for eg the flu. The symptoms of the flu and of COVID-19 are pretty similar, so it could have been the flu, not COVID-19, that killed the person, but their deaths will now be attributed to COVID-19.

7. It is not uncommon for deaths to be classified as COVID-19 deaths even in the absence of a positive COVID-19 death. In the US and UK it is now accepted practice to count a death as a COVID-19 death even if there is no positive test result as long as the doctor makes a subjective judgment that it likely was COVID-19 that caused the death. There is a lot of room for subjectivity in that determination.

8. The extent of the COVID-19 problem is not put into context by comparing it to other problems. So while e.g. 500 COVID-19 deaths may sound very scary, if in that same period 10,000 people died from the flu it may seem less so.

9. Exponential growth rates are assumed to continue at that rate instead of quickly levelling off.

10. An excessive reliance on models that are only as good as their assumptions.

11. Relying on experts whose past pandemic predictions were way off.

12. Governments taking enormously far reaching steps to fight the problem give rise to a “Well, they wouldn’t take such extreme measures if the problem weren’t that extreme” attitude.

13. Other countries —such as South Korea, Taiwan, Japan and Singapore— appear to have been quite successful in containing the epidemic without resorting to the kinds of draconian measures we see Western governments taking.



9 comments:

  1. 2. The situation is changing daily. Isn't it obvious that planning is extremely important? This is like suggesting that reporting should focus only on the visible part of the iceberg....
    3a. These are facts that contribute to the problem, i.e. the system being overwhelmed, and thus are evidence for the threat.
    4. I don't understand this point. Are you saying that it would be less scary if we didn't know what is going on?
    5. It seems obvious that COVID-19 contributed to these deaths... maybe they should count as some fraction of a death in predictive models?
    6. Again, this is like saying everything's not so bad right now, so we're probably fine... COVID-19 is much more contagious than the flu, and the number of cases is increasing.
    7. With business as usual, wouldnt they level off close to 100% infection? Why would you assume they would quickly level off?
    8. What do you suggest?
    9. Do you have any evidence for this point? So far the predictions seem to have been proven fairly accurate as far as I can tell.
    10. This point doesn't seem to support your thesis. The opposite would also hold true. And if citizens don't adhere to the government steps, they won't work.
    11. This point deserves elaboration... It's by far the best one.

    ReplyDelete
  2. Keep in mind throughout that my central thesis is not that it's not a serous problem or that we should not do anything. I exlicitly say that it is and that we should. At least some of your criticism seems to be based on the assumption that I'm denying either or both.



    "2. The situation is changing daily. Isn't it obvious that planning is extremely important? This is like suggesting that reporting should focus only on the visible part of the iceberg...."
    "
    Of course planning is extremely important. Again, my point is that we may be getting an inflated sense of the threat. For example, we see tons of headlines about insufficient capacity at hospitals. It isn't until you read the articles, usually at least a few paragraphs in, that you find out these are projections instead of current reality. So the average reader may think 'Oh, this is already a big problem and it's only going to get worse' even though it is not currently a problem (some exceptions apply). Moreover, the projections may prove to be accurate but the probability that they do is necessarily lower than the probabiity of whatever the current situation is.

    So readers get an inflated sense of the problem.

    Might be nteresting to do an opinion poll and ask people how great they think the current capacity problem at hospitals is.

    the iceberg analogy is problematic. An iceberg is already pjysically there, in its entirety. It's just that we only see the tip. In the case of these prjections only current reality is already there, the rest of the iceberg is projected, and the projection is only as good as the assumptions the model is based on.

    ReplyDelete
    Replies
    1. 3a. These are facts that contribute to the problem, i.e. the system being overwhelmed, and thus are evidence for the threat."

      Again, I am not denying that there is a threat. I'm only saying we may be getting an inflated sense of it. If the reader is not aware that hospitals often run at near capacity or that in flu season it is quite common for them to be over capacity then if they read in the newspaper that right now hospitals are having capacity problems they will be more inclined to assume that this is because of Covid-19, creating an *inflated* sense of the threat. Numerically: If readers assume a hospital is normally at 75% capacity and now they hear that it is overloaded, then they will assume 25+% of the capacity is used for Covid-19. If they knew that the hospital is normally at 90% capacity and now they hear that it is overloaded, they will assume 10+% is used for Covid-19. And so the problem is inflated in their minds.



      "4. I don't understand this point. Are you saying that it would be less scary if we didn't know what is going on?"
      "
      It would be very odd if that is what I meant. Suppose only people with mild to severe symptoms will tend to be tested: 5,000 people are tested and 4,000 test positive. So people who tested positive tend to be pretty ill. Now assume there is a big surge in testing so that a lot more people with no symtoms are tested. The number of tests is now 100,000 and the number of positive tests is now 10,000. So that's a huge increase from 5,000 to 10,000. But because the scope of testing is now expanded there will necessarily be a lot more people who tested positive in this new batch but who are asymptomatic or have only very mild symptoms. But if we still associate a positive test with pretty serious illness then an increase from 5,000 to 10,000 is reason to panic. Yet if we realize that the increase in positive tests is mainly due to people who are asymptomaic or have only very mild symptoms now getting tested, there is much less reason to panic. The number of people seriously ill with Covid-19 will be much smaller than the increase in positive tests suggests.

      Delete
    2. "5. It seems obvious that COVID-19 contributed to these deaths... maybe they should count as some fraction of a death in predictive models?"
      "
      On what basis do you say that it seems obvious?

      What percentage of people who test positive are asymptomatic or have only very mild symptoms in a normal population?
      That percentage is probably lower in a population with serious underlying illnesses but it's still not a given that the virus contributed to their deaths.

      Also, suppose somebody who was already very likely to die within a year now gets infected, the virus causes symptoms and those symptoms at least to some extent hasten that person's death. That is bad but it's obviously a very different situation than an otherwise healthy person getting infected and dying solely from the illness caused by the virus. To the extent that people think the latter type of case is the norm they are getting a very inflated sense of the threat.

      Take for example Italy. Italian authorities say that 99% of people who died and tested positive for Covid had underlying conditions. https://www.bloomberg.com/news/articles/2020-03-18/99-of-those-who-died-from-virus-had-other-illness-italy-says
      The Imperial College prof whose models are key behind a lot of governments' Covid decision making says that maybe half to two thirds of the people who died from Covid would have died within one year anyway.

      Still bad, but considerably less bad than if this was about otherwise healthy people.

      Moreover, it seems likely that at least some people who now die from Covid would otherwise have died from e.g. the flu. Dunno how high that percentage is though.

      Also, "The average age of those who’ve died from the virus in Italy is 79.5. As of March 17, 17 people under 50 had died from the disease. All of Italy’s victims under 40 have been males with serious existing medical conditions."

      Your 'some fraction of a death' idea is interesting. Not sure how to conceptualize it but i think there is a there there.

      Delete
    3. 6. Again, this is like saying everything's not so bad right now, so we're probably fine... COVID-19 is much more contagious than the flu, and the number of cases is increasing."


      It's not like saying that. if you insist it is, please make the comparison as explicit as possible because I'm not seeing it.
      re covid being much more contagious than the flu. Please point to the specific data that you base this on. I don't mean, refer to a source that says this, point to the specific data. I don't think at this point we have enough data to make a confident statement like that.

      For example, if we don't know what percentage of the population has or has had the virus without developing any symptoms or with only mild symptoms it will be very difficult to determine how contagious it is.

      One of the best things that should be done now is test a small random sample of the population for 1) the virus 2) antibodies for the virus. That would vastly improve our understanding of the nature and extent of the pandemic, how



      "7. With business as usual, wouldn't they level off close to 100% infection? Why would you assume they would quickly level off?"

      Nobody is advocating business as usual.

      Why I would assume it would quickly level off? WHat viruses have infected 100% of a population?

      Before the flu vaccine, what percentage of a population would typically get the flu?

      What percentage of the population typically gets new strains of the flu that the vaccine does not protect against yet?

      Delete

    4. "8. What do you suggest?"

      To be careful when working with models and keep in mind that they are only as good as their assumptions. Don't take them as gospel.

      Other than that, maybe as much as possible continuously compare the model's predictions to empirical reality and see where they are accurate, where not, and when sensible, adjust, and adjust one's confidence in them.

      "9. Do you have any evidence for this point? So far the predictions seem to have been proven fairly accurate as far as I can tell."

      this admittedly was ambiguous but I was talking here about their predictions for *past* pandemics. And here's some evidence for that:
      outline.com/http://www.telegraph.co.uk/news/2020/03/28/neil-ferguson-scientist-convinced-boris-johnson-uk-coronavirus-lockdown-criticised/

      re predictions for the current pandemic: What do you base your assessment that the predictions have been proven fairly accurate so far on?


      "10. This point doesn't seem to support your thesis. The opposite would also hold true. And if citizens don't adhere to the government steps, they won't work."

      I don't understand what you mean by your first 2 sentences.
      re the third sentence: that's trivially true but i don't understand how it undermines my point, so I'm probably misunderstanding what you mean.

      "11. This point deserves elaboration... It's by far the best one."

      and I'll add: What actual data do we have to conclude that shutting down large parts of the economy is an effective way to contain the virus? I understand the theory behind lockdowns (but even there there are some counterarguments) but do we actually have data convincingly showing lockdowns are effective (necessary and/or sufficient)?

      Moreover, the lockdown is not intended to prevent people dying from the virus, only to prevent the additional deaths due to lack of medical care if hospitals become overloaded. Other than that, the lockdown just distributes virus deaths over a longer period, unless a vaccine and/or cure is found in record time (and there is some hope for an effective treatment)

      As concretely as possible, what would convince you that the current lockdown is a disproportionate (or maybe even ineffective) response to the problem?

      For myself, I very very much would like to see analyses comparing the realistic economic, political, social, institutional, medical and psychological effects of a lockdown, and comparing it to realistic effects of the virus if no lockdown is imposed but other sensible measures are taken.

      Incredibly, nobody has ever attempted such an analysis.

      Delete
  3. So your central thesis seems to be that the threat of the Corona virus is or may be inflated. You haven't really expanded your thesis beyond that. If you want to argue that people in general or in a specific country should be behaving differently somehow, I think that is probably irresponsible. I don't think that's your intention, but please correct me if I'm wrong. If you want to argue that governments' responses should be different because their perception of the threat is inflated, then I don't think your responses to my criticisms of points 2, 3a, 4 help that case.

    Before going much further (I'm already at the limit of my bandwidth, to be honest) perhaps we should clarify what your intentions in writing this post are.

    A few brief thoughts about your responses:

    2, 3a, 4: It seems like your argument boils down to assuming general media illiteracy.

    5: I think we are more aligned on this point now. In your initial post you seem to be stating that those deaths where the patients had another condition besides COVID-19 shouldn't be counted at all. I hope you agree that doesn't make sense.

    6: I don't have the bandwidth nor the expertise to find and parse the data. I'll revise my belief to COVID-19 being at least as contagious as your average flu with no vaccines.

    7: So we (you and I) just don't know. I give the benefit of the doubt to the modelers here. I'm assuming they've looked at many different models with different assumptions and have some ideas as to their likelihoods. If you have any evidence to the contrary, this might be a valid point.

    8: Sure. Again, do you have any evidence that this is not being done? Seems best to give the benefit of the doubt to those in charge here.

    9: I don't have the bandwidth to give a satisfying answer.

    10: I guess to rephrase my objection to your point... at best, isn't it just a truism? Perhaps clarifying and/or extending your thesis would give this point some signficance.

    11 (your additions): 1st paragraph.... Obviously we are trying to limit the spread of the contagion, which involves reducing points of contact, or chances for infection. Do you have any data or evidence? 2nd paragraph... The point is that we want to reduce the burden on the healthcare system as much as possible... to reduce total deaths.

    As to your question about what would convince me "that the current lockdown is a disproportionate (or maybe even ineffective) response to the problem"... perhaps an alternative proposal would be the most important thing, backed by experts and data and analysis that I likely wouldn't realistically have the bandwidth or likely the ability to understand. In short, concensus.

    ReplyDelete
    Replies
    1. "So your central thesis seems to be that the threat of the Corona virus is or may be inflated. You haven't really expanded your thesis beyond that"

      Because that is the only point I set out to make in this post, as the title and the introduction indicate. You can't blame me for failing to argue for a point I wasn't trying to make.

      Delete
    2. I don't think continuing the discussion is going to be very productive and I get the sense you feel likewise, so I suggest we leave it at this. Maybe it will work better in person some time.

      Delete