How Many Years of Life Does the Average Covid-19 Victim Lose?

An exchange with the lead author of a study.

|The Volokh Conspiracy |

A few weeks ago, the Wall Street Journal reported on a study suggesting that the average person who dies of Covid-19 loses ten years of life. I looked up the study, and left the following comment on the study's site (which, to the author's credit, they have up to get instant peer review), after noticing that the author's seem to assume that the victims were previously as healthy as their demographic average [including comorbidities]:

Two people who are coded with the same disease could be in vastly different circumstances. We know the virus has taken a huge toll on nursing homes.  An 82 year old with heart disease who lives in a nursing home is not similarly-situated, life expectancy-wise, to an 82 year old who is otherwise doing well and is self-sufficient. The former would assumedly be much more likely to succumb to Covid-19 than the latter. Similarly, "otherwise-healthy" people who succumb to Covid-19 can be expected to, on average, be more likely to have an undiagnosed health issue than those who don't. Is that taken into account? If neither of these are taken into account, the effect on life expectancy must be reduced.

Now, I see you've responded [to another commenter[ that this should NOT have a major effect on life expectancy. I don't see how you can be so confident. A *huge* percentage of deaths, wildly disproportionate, have been in nursing ("care") homes… You simply can't compare an otherwise healthy 82 year old with heart disease to someone whose heart disease so enfeebles him or her that they need to be in a nursing home.

A day later, another commenter wrote:

I'm perplexed by this study. How can it be assumed that the Covid victims would have lived the average life expectancy unless there's no or minimal standard deviation around that average? Wouldn't it be more compelling to compare to the minimum life expectancy of each cohort? Otherwise, you are implicitly assuming that the people who are dying are more or less representative of the average, which seems like a major assumption that, if untrue, would render your conclusions pretty useless. I hope I'm missing something here because it would seem far more intuitive to assume that people who are dying are the most vulnerable of their respective cohorts.

A few days later, lead author David McAllister responded:

In response to Jason Bloomberg and David Bernstein.

Our work was a response to the assertion has been that "because those dying are older and have lots of comorbidity, they probably don't have to live". I think JB and DB may be making a different statement that "notwithstanding the fact that the average life expectancy is still quite long among older people with comorbidity, those dying from COVID-19 are likely atypical compared to the average among older people with comorbidity".

I think we are talking here about residual confounding, i.e., after you take into account the known/measured variables, are there remaining differences between patients on which we estimated life expectancy (the general community in Wales) and those dying of COVID-19 in the Italian data.

I think one has two options with residual confounding. Either to state this as an assumption/limitation and/or try and model it in some kind of sensitivity analysis.

Professor Andy Briggs effectively does the latter (https://avalonecon.com/estimating-qaly-losses-associated-with-deaths-in-hospital-covid-19/) looking at the effect of quite large multipliers on life expectancy, implemented via an excel tool. This would allow the commentators or others to explore the impact of different mortality rate ratios based on different assumptions as to the degree of residual confounding.

We have taken the former approach. As we are not aware of any empirical evidence to provide us with an estimate for the magnitude of the residual confounding due to unmeasured characteristics (e.g. frailty, functional limitation).

This is because, in order to make the assertion that those dying from COVID19 are atypical of their fellows who are similar in terms of age, sex and comorbidity we would argue that empirical evidence to support that claim is needed. Not least because, although we cannot know how strong they are, there may be selection pressures in the opposite direction. For example, someone with relatively mild COPD might go food shopping themselves, whereas someone with more severe disease might have someone else shop for them, thereby reducing their infection risk. Since the risk of death is the product of the risk of infection and the case fatality, this mechanism would tend to select for less severe COPD among those dying from COVID-19.

We argue that additional data, ideally on functional limitations (e.g. able to walk to shops, able to walk up stairs) and frailty measures (e.g. grip strength, lung capacity, six-minute walking distance) should be obtained to allow us to estimate the YLL more accurately using more empirical evidence.

Nonetheless, we think that this reasoning should not be applied to care home residents. Our results came out before the large numbers who were dying in care homes became apparent and this was not the focus of our work. Instead we agree that we should estimate mortality (and YLL) in care homes separately. Importantly, care home residents are a well-defined population so the task of estimating life expectancy in this group should be achievable in most settings.

Some readers may be wondering what a law professor is doing commenting on scientific papers, but I've been researching and writing about the reliability of scientific evidence for over thirty years, so I'm not a complete noob. And of course it goes without saying that all lives are valuable, even if scientists and policy-makers have good reasons for wanting to know how Covid-19 is affecting victims' life spans.

Advertisement

NEXT: Robert E. Lee Becomes Roberto Eduardo Leon: An Affirmative Action Controversy from 1979

Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of Reason.com or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Report abuses.

  1. When I first heard of this study, I mistakenly believed it was an estimate of how much the Wuhan virus shortened the lives of those who had survived it — one never fully recovers from things like pneumonia.

    The other study that needs to be done is the morbidity and mortality caused by these fascist shutdowns. Suicide is up, domestic violence is up, drug abuse is up, while treatment for the top two killers (Cardiac & Cancer) is way down.

    This *will* shorten the collective lifespan — I don’t know how much but as the suicides and drug addictionss tend to be young adults, the years lost are significant.

    No one seems to care about this — perhaps because it isn’t sexy and perhaps because there isn’t research money in discovering just how wrong our purported “experts” actually are, but dead is dead and people are dying from all of this.

    At least at “John Muir Medical Center in the East Bay region of the San Francisco Bay” they are seeing way more suicide deaths than Wuhan ones, https://thefederalist.com/2020/05/22/california-doctor-calls-for-end-to-lockdowns-over-rising-suicides/

    Other examples abound, but no one is talking about it.

    1. I don’t disagree with your post, with one exception.
      Quote “one never fully recovers from things like pneumonia”.
      I had pneumonia when I was 11, I am now 64, and I assure you I am fully recovered.

      1. It might be interesting to track people and their average lifespan in that case.

    2. Are you quite sure that you know what the word “fascist” means?

  2. One thing I greatly appreciate about Prof. Bernstein is that he seems to have a stronger commitment to truth than to “scoring points” for his preferences or (“his side’s” preferences).

    Does he have preferences/biases? Yes. (We all do.) Nevertheless he raises reasonable questions in a civil way.

    One of the biggest dangers of “deplatforming”, “cancelling”, and “blocking” is how it can lead to blinkered groupthink. Regardless of how it impacts any political “team” or “clan”, surely it is more important that we have a good understanding of the effects and mechanisms behind this pandemic.

    1. Well said. Sometimes Prof Bernstein really gets my goat, but only for the topics, never for petty sniping or insults or trying to hide his bias. I’d much rather read his articles than a lot of others which try to pretend to be neutral.

  3. “…the average person who dies of Covid-19 loses tens years of life….”

    This is an unfortunate typo. You obviously mean “10 years of life” OR you mean “tens of years of life.” A huge difference, and nothing in the rest of your OP tells us which one you meant to write. Of course, my first step was to go to the linked WSJ article, but it’s behind a paywall, so it was impossible to figure things out that way.

    David, can you clean up the typo? Or post here to tell us which is correct? (Posting here is probably less easy than merely correcting the typo, I suspect.)

  4. If I wanted to guess how many years of life a nursing patient lost when he died of Covid-19, I would not just compare him to the whole “cohort” determined by his age. I would compare him to data from nursing home patients in the area, or better yet, patients in that particular nursing home. That way I would pick up the effects of the patient’s socioeconomic class, which predicts life expectancy better than mere age, and possibly also his prior health history (and that of other patients in the same home) which will predict it even better.

    1. I would compare mortality statistics over the past 10 years, preferably for that home but you won’t have statistically significant data.

      My point is that a lot of the nursing home patients were statistically likely to die of *something* this winter, and hence the relevant data is how much the TOTAL death rate this year is increased over prior years.

    2. I think the authors agree with you.

      Nonetheless, we think that this reasoning should not be applied to care home residents. Our results came out before the large numbers who were dying in care homes became apparent and this was not the focus of our work. Instead we agree that we should estimate mortality (and YLL) in care homes separately. Importantly, care home residents are a well-defined population so the task of estimating life expectancy in this group should be achievable in most settings.

      Emphasis added.

      1. This may be the first time I’ve influenced the course of an in-progress study, so that’s kinda cool.

        1. Well, yes it is, but I think your comment that “the effect on life expectancy must be reduced” is not proven. The authors make a reasonable point that it may be the case that healthier individuals are somewhat more likely to get infected, by virtue of their own less cautious behavior, and that this might outweigh the effect you mention.

          Best to wait for the data.

          1. That’s what makes it “an exchange.”

            1. True.

  5. Insurance man here.

    Mortality tables don’t state averages; they state medians. For any given age, there will be a median mortality; that is half will live longer and half will live less. I am not sure how the study goes from medians to averages. Maybe that is a typo.

    Mortality for persons in assisted living, i.e. care homes is indeed different than for persons not in assisted living. Standard life insurance mortality tables do not apply to persons in assisted living. Providers of long term care insurance utilize different tables to figure premiums for prospective insureds. Medicaid also keeps data on mortality for persons in assisted living; because Medicaid provides coverage for assisted living after a recipient has completed the full Medicaid spend down. Medicaid recipients typically are poorer than the insureds who obtain private long term care insurance and likely to be in poorer health than recipients who have private coverage. So the mortality of Medicaid recipients might be shorter than for private recipients.

    Persons in assisted living have much shorter lifespans than their peers living independently. Why do people in assisted living not live as long, regardless of age? People usually go into assisted living once they are unable to perform one or more of the standard activities of daily living: eating, bathing, getting dressed, toileting, transferring, and continence. Being able to independently perform these activities are critical to avoiding illness. Once you can’t do them you become vastly more susceptible to illness and complications from illness. Generally, crudely, just depending, once a person goes into an assisted living facility, the median life expectancy is about 18 months, regardless of age upon entry.

    1. “Mortality tables don’t state averages; they state medians”

      Ummmm — medians *are* averages….

      There are three forms of statistical averages:

      1: Mean Average — Add them up and divide. This is what most people think of as “average” but it can be skewed by one or two outliers. For example, the average age in a Kindergarten class of 10 five-year-olds and a 50-year-old teacher is 9 years old.

      2: Median Average — half above and half below. For example, we use “median” household income and not “mean” household income because those with very high or low incomes would skew the statistics.

      3: Modal Average — which number appears the most. (Multi-modal data has more than one mode.)

      1. Thanks. Insurance mortality tables speak in terms of the median average. It doesn’t change anything else about what I said. In fact, it makes reading the study more challenging since the author doesn’t state what kind of average he is referring to.

        1. Exactly — and what infuriates me is there are standards fir statistical research — the dry boring stuff that every Doctoral Candidate has to learn and which nobody ever questioned before — which are being totally ignored here….

          1. I mean the larger “here” — in all the hysteria about what I insist on calling the Wuhan Virus.

      2. I have to say that the first of those, the mean, is the only one I can recall seeing referred to as an “average”, with the other two simply called the “median” and the “mode”.

        A quick search proves to me that your usage isn’t unheard of, (What usage would be?) but it’s not typical.

        1. When learning statistical methods as an undergrad and then grad student; we students were told the old joke of the statistician who drowned in a lake with an ‘average’ depth of 12 inches.

          (And, of course, there is the even-more-famous quote of “lies, damned lies, and statistics.”)

          1. Ask me sometime about the Harvard students who would have drowned in not that much more water than that, if I hadn’t “encouraged” them to stand up & walk ashore.

            It sorta seemed like the Christian thing to do at the time…..

          2. That’s the way the world goes round
            One day you’re up, the next your down
            Its a half an inch of water and ya think your gonna drown
            That’s the way the world goes round

            John Prine

    2. Does median life expectancy differ significantly from mean life expectancy?

      Surprising, to me anyway, if so.

      1. bernard11, the answer is: How big is the group you are comparing?

        For large groups (like 1MM+), there will be no meaningful difference; the 50% point (median) will become indistinguishable from the average.

        1. That’s actually not true, and the two can be quite distinguishable. Here’s the table from the social security administration.

          Mean life expectancy is 76 (Male, US). But Median life expectancy is 80 (Male, US).

          The problem lies in a “high” in one end of the table (in this case, deaths under 1) which end up skewing the means.

          You also see this effect when comparing mean versus median income.

          https://obliviousinvestor.com/mean-vs-median-life-expectancy-for-retirement-planning/

          1. A.L.,

            That’s true but the group we considering here doesn’t include infants, so the infant mortality rate isn’t really an important factor.

            Are there differences for adults is what I really want to know.

            XY,

            I don’t think that’s true in general, since populations can easily be skewed in some metrics. Still, for something like life expectancy you are certainly correct, once you get past the infant mortality issue A.L. raised.

            1. If you keep reading the linked article, the mean and median life expectancy are still different.

              “For instance, for a 60 year old male:
              Mean age at death is 81.5 years,
              Median age at death is 82.5 years, and
              Mode at age death is 86 years.

              And for a 60 year old female:
              Mean age at death is 84.5 years,
              Median age at death is 86 years, and
              Mode age at death is 89 years.”

              This is due to the non-symmetrical curve, which tails on the younger end, then has a sharper drop off on the older end.

              1. OK. Thanks.

          2. Iirc a big chunk of improvements from ancient “average life expectancy is 35” came from improvements in baby deaths. If you made it past that, and could avoid accidents and infections and starvation, the length of life was not so different.

  6. It is worth noting that one of the preconditions pointing toward deadly Covid-19 outcomes—maybe the most statistically influential precondition—is age itself. Pretty obviously, that one can’t be an indicator of disease-related shorter life expectancy for an old person, lest the entire actuarial enterprise go out the window.

    So during the pandemic, whatever does point to shorter life for an old person must be some factor other than age. Probably it would make sense to concentrate on other co-morbidity factors themselves, and ask first what is the life expectancy of a person with that factor at that age, if he never gets Covid-19. If the factor shortening life expectancy is something like COPD, or heart failure—conditions promising shorter life expectancy on their own—then, as with age, that much-shorter life expectancy is already accounted for in the life expectancy table. Such tables are designed to include mortality effects of well-known deadly conditions. Covid-19 thus has little room to shorten life expectancy in such cases of imminently deadly preconditions.

    On the other hand, if the the comorbidity is something which does not normally much affect life expectancy—auto-immune arthritis, for example—but which does affect Covid-19 survival—then the loss of expected life cannot be attributed reasonably to the co-morbidity factor. It has to be attributed to Covid-19 itself. Absent Covid-19, that person previously had a 50% chance to outlive his life expectancy, despite his condition. If more such preconditioned patients die short of their expected time, then Covid-19, not the preconditions, is the cause.

    In short, in either class, whatever Covid-19 does to measurably reduce life expectancy in older people cannot be much on account of preexisting conditions. Conditions which tend to kill people promptly anyway will not much budge the life-expectancy needle if Covid-19 intervenes. That early death was already mostly-accounted-for in the expectancy.

    Preconditions which do not by themselves much affect life expectancy are different. When those do predispose to deadly Covid-19 outcomes, the life lost is rightly attributed to Covid-19, not to the precondition.

    Thus, the argument that for policy purposes, expected years of life lost to Covid-19 can rightly be discounted, because the preconditions are the real culprit, is logically suspect. That raises the question why that argument is so commonly made, and so heatedly insisted upon.

  7. A 57 year old individual I am aquainted with showed up at the emergency room and died within an hour. Apparant heart attack – but he tested positive for covid19, so he will go down as dying of covid in the medical stats.

    How many years of life did he lose to covid ? maybe 10 minutes?

    1. The first person to die of Covid-19 in Humboldt County was 97 years old. She was in a nursing home for the usual reasons, and we can be sure that decades of a long, productive life were stolen from her.

    2. There’s probably no way to tell in the case of your acquaintance but it’s potentially decades since heart attacks seem to be one of the symptoms.

      How big an issue that is will probably be something that epidemiologists will be sorting out for the next few years.

      1. Ultimately we are going to have to look at total (estimated) excess deaths to get a measure of the effect.

        1. And that is only a guess. Moreover, there are those who die of heart attack who likely would not have died except that they were afraid to go to the hospital. We are unlikely to know the death toll from sequestration and the attendant generation of fear especially among those over 60 (or so).

          1. Well, it’s an estimate based on relevant available data. You can call it a guess, I suppose, but it’s not going to be a number pulled out of someone’s rear end.

            Presumably the estimate will have some sort of confidence intervals.

          2. Don,
            I expect the number of people who say, “Wow, I’m having a heart attack!” and also “I’m so afraid of Covid that I won’t go to the ER…I prefer to face my heart attack at home.” is close to zero. Sounds like an insane and moronic calculation to me. (I, sadly, have met too many really dumb people for me to feel confident saying that the number of these cases is actually zero…there will always be *some* people who do the most foolish thing, alas.)

            1. Many people don’t say “Wow, I’m having a heart attack”. They say “Huh, my chest hurts, and I’m out of breath, what do you think honey”.

            2. One of many news reports to the contrary.

              1. Appendicitis is not a heart attack.

                Nobody who is having a heart attack thinks “I’ll just wait it out at home and take my chances because I might die if I go to the hospital.”

                Try a bit harder not to be a moron.

                1. Not at all implausible that such an individual might not have experienced the heart attack but for the covid-19 infection and gone on to live his otherwise expected life span. This extremely perverse virus can cause catastrophic neurologic, renal, and cardiac complications as well as the more common ling injuries.

                  1. Is thought to cause — not “can.”
                    We don’t know that with certainty yet.

                    1. b@Dr. Ed Covid-19 patients without risk factors for major vessel strokes or acute renal failure develop these surprising condition in the face of this particular viral infection. But you claim we only [i]think[/i] the covid-19 virus may cause these intercurrent problems (e.g., large vessel strokes, acute renal failure, abnormal blood clotting), we aren’t able to say it [i]can{/i] cause them, “we don’t know that with certainty yet”?!

                      With all due respect, you don’t know what you are talking about.

                2. Wrong. People certainly do fail to go to the hospital with definite symptoms and signs of heart attacks and strokes for various bad reasons. This is why so much time, effort, and money goes into efforts to educate the public about these very time-sensitive health events.

                3. OK, here’s one for a heart attack.

                  And it’s not just that one patient: “Parwani said as at other hospitals, she’s observed about a 40% drop in heart attack patients coming to emergency rooms.”

                  This is one of many such articles a cursory google search will find.

                4. The decline in heart attacks going to the hospital is probably due to increased sedentary lifestyle during lockdown. While this could be worse in the long run, it’s beneficial in the short by less strain on the heart doing things.

                  It’s like a mild winter where fewer fatsos have heart attacks shoveling snow, x everything humans do daily.

                  The idea people are stupidly toughing it out always seemed like a rationalization pulled out of a magician’s hat to me. It also has the advantage we can enjoy thinking about it by feeling superior to those idiots!

                  1. A quote from the previously linked article: ‘Critical care cardiologist Dr. Purvi Parwani of Loma Linda University Health said, “We have some patients in-house that presented late because they were truly concerned about getting the COVID infection.” … But it’s not just heart patients who are avoiding going to the hospital.

                    At USC Verdugo Hills Hospital, overall ER patient volume is down 40% to 50%. The facility is seeing COVID-19 related cases, but emergency medicine specialist Dr. David Tashman said, “All the rest of the chest pains, abdominal pains and other complaints that we’re used to seeing on an everyday basis, really sort of, well, we’re all a little bit dumbfounded. Where did these patients go?”‘

                    These articles are easy to find. Here’s one discussing strokes inter alia: “Since state and county officials issued orders last month largely instructing residents to remain at home, the number of stroke cases coming to the hospital has dropped by half, Brown said.”

                    And the docs know the reason people are delaying: “Asked why she had waited, the woman told doctors she had been afraid she’d be exposed to the coronavirus at the hospital, Brown said.”

                    Again, even a casual search will find many similar articles. Out local paper has had similar local stories.

            3. Bullbleep.

              And it’s not just the actual heart ATTACKS as much as the arrhythmias and other stuff which a competent MD can both see and treat BEFORE THERE **IS** A HEART ATTACK.

              1. You really should stop with the medical opinionating. Myocardial infarctions (heart attacks) not infrequently cause cardiac arrhthymias, some of those an immediate cause of death (venticular fibrillation); cardiac arrhythmias are rarely the cause of myocardial infarctions.

                1. Well then we apparently are wasting a lot of money sending people to cardiologists then. Is that what you are saying?

                  1. No. Again, you presume to speak about medical matters you are not even minimally informed about.

                  2. No, I’m saying nothing of the sort.

                    Again, you presume to speak about medical matters you are clearly ignorant of.

            4. My mother-in-law literally did not know when she had a heart attack. After several days of feeling weak, she finally scheduled a doctor’s appointment, which led to tests that finally revealed damage from a heart attack at least a week earlier. Michigan has banned most non-emergency medical care, so it’s quite likely that if this happened now, she would not have been able to schedule that appointment.

              OTOH, there was really nothing that could be done about her damaged heart, short of a transplant – and she was too old for that. The doctors correctly predicted that her heart would finally fail within six months. All that medicine could provide was the warning so we could prepare to lose her. She went on hospice care, and my wife moved into her home to take care of her.

              But surely _some_ of all the check-ups, non-emergency surgeries, etc., that are temporarily banned would actually matter.

        2. “Ultimately we are going to have to look at total (estimated) excess deaths to get a measure of the effect.”

          That will still take a lot of work to tease out the covid deaths. For example, the things people are doing to avoid covid – washing their hands, avoiding crowds, etc – are also effective against influenza, and you’d have to account for that to do an accounting of flu vs. covid deaths.

          1. That’s true, but maybe whatever lives are saved that way count against lives lost to covid.

            IOW, if we want to measure total lives lost due to covid, it seems reasonable to count lives saved against that total.

            Suppose, very hypothetically, that getting covid reduced the chance of having a heart attack. Then in computing deaths due to covid we would surely take into account the reduced number of heart attacks. Indeed, if the reduction were big enough it might turn out that the virus was beneficial, by substituting nine virus deaths for ten heart attack deaths.

            1. How do you imagine covid-19 would reduce the chances of another fatal illness or postpone one? I can’t.

              1. “By the way, did you know that your blood pressure is 210/110?
                You really ought to see someone about that…”

                You can’t imagine that happening?

              2. How do you imagine covid-19 would reduce the chances of another fatal illness or postpone one? I can’t.

                I don’t really imagine that. It was purely hypothetical.

    3. Not at all implausible that such an individual might not have experienced the heart attack but for the covid-19 infection and gone on to live his otherwise expected life span. This extremely perverse virus can cause catastrophic neurologic, renal, and cardiac complications as well as the more common ling injuries.

      1. It probably also causes male pattern baldness….

        Coincidence is not always causation — it can be but often isn’t.

        1. OK, we are convinced. You are far too ignorant about what you presume to speak to waste time with.

  8. On the last point, I will say that while I respect that Prof. Bernstein has done some work in the area, one of the worst things about discourse about the pandemic is that almost all of the loud and prominent voices have been from people who have no degree in epidemiology, have done no academic work in the field, and have never worked in the field.

    Everyone who comments on other things seems to think they are an expert on epidemiology. Meanwhile, the people who we should be listened to have fewer platforms. So there’s way too much Nate Silver, who doesn’t know crap about the subject but thinks he does because he knows about statistics, and Richard Epstein (at the beginning) making a fool of himself, and just all sorts of people who normally comment on politics or the law or whatever who suddenly think they understand this incredibly complex field.

    I wish all of us would take a step back and let actual experts in this field talk. People do not need to pretend they understand everything in every field.

    1. wrt Nate Silver….his statistical modeling isn’t so great. GIGO.

      1. Silver’s political models and analyses are quite good. He has done almost no modeling of COVID-19, so I don’t know what Dilan is referring to.

        1. He’s been a nonstop font of commentary on the epidemiology of the coronavirus since the very beginning of the epidemic in the US.

          1. He keeps what various models predict as one of his headlines. He has posted on the difficulty of COVID-19 modeling. Has has interviewed the creator of the IHME model. He has pointed out that the number of COVID-19 cases is not a good metric because of inconsistent testing regimens.

            I find none of the above problematic.

            1. Josh,
              I agree. The one thing Silver has said over-and-over is that we will not have a good sense of this until months and years have gone by. (And, of course, that we should do the best we can with the data we currently have, in the meantime.)

              I’ve found his informed speculation to be measured and reasonable.

    2. Too damned many actual experts are no such thing when their biases and ideology are at stake. Michael Mann is an excellent example.

      I will never defer to experts. I will listen to them, I will give them an initial benefit of the doubt, I will give their evidence more weight. But they became experts by proving the previous experts wrong, and their expertise will eventually be proven stale too.

      1. Á àß — I suppose there is room to claim expertise by proving previous experts wrong. That doesn’t strike me as the usual process.

        The first step in the usual process seems to be to recognize that expertise is not an individual attribute, or even something inherent in the expert. It is instead about participating in a shared body of knowledge, managed, analyzed, and communicated by agreed-upon shared methods. Compared to thinking every complicated problem through on your own, there are big advantages to relying on shared knowledge which has been critiqued by other especially-interested people.

        Sometime the experts will all go wrong together. When that happens, they will be far more likely than the lone-wolf thinker to include among their number the one person who is right to help them correct themselves promptly.

        Autodidacts typically earn their equivocal reputations. There have been brilliant exceptions, but hoping to become one of those exceptions is a less reliable means to insight than expertise. Quite often, group-knowledge expertise even outshines individual brilliance.

        1. The first step in the usual process seems to be to recognize that expertise is not an individual attribute, or even something inherent in the expert. It is instead about participating in a shared body of knowledge, managed, analyzed, and communicated by agreed-upon shared methods.

          Right. And alongside of this, add “experience” as well.

          As a lawyer there’s a whole bunch of stuff I learned both in law school and then in the practice of law that it would be very difficult for even a skilled non-lawyer commentator (such as the late Anthony Lewis) to match. So while obviously participation in the shared body of knowledge is a big part of it, training and work experience are also a big part of it as well.

          A whole bunch of vocal Americans really do believe that they can just start reading an epidemiological study, or even worse, someone’s second hand report of such a study, or look at a couple of graphs, and boom, they understand the coronavirus pandemic. It’s crazy. People literally go to school for almost a decade and then work in labs and publish papers for decades more to become expert in this stuff.

          1. Dilan, of course you are right about all of that. Thank you for correcting my oversight on experience—a very important part of expertise, which the experts’ practice of sharing knowledge enhances.

    3. It all depends on what topic is being discussed. If it’s how much should we lockdown and when/how to re-open, the last people I want to hear from are epidemiologists. They are hammers, and normal life is a nail. Even restricting oneself to epidemiology, I find many epidemiologists not credible. There has been a huge amount of uncertainty about this virus, yet the vast majority of “experts” act as if they have more concrete information than actually exists, including epidemiologists.

    4. Everyone who comments on other things seems to think they are an expert on epidemiology

      I don’t know a damned thing about epidemiology. However, I am a former professional statistician, and I can tell you with absolute certainty that majority of epidemiologists that are making predictions do not understand the statistics they are using.
      Probably, most of them are making errors in the their models and in their analyses.
      Beyond that, modelling usually uses custom programming – and that requires skilled programmers to produce programs that perform the correct calculations that the math relies on.

      The Imperial London model is an excellent example: The author released the code for his models to Microsoft, who spent a month a dew million dollars cleaning it up, then released the source code.
      It was massively buggy, to the point where it could not produce identical runs even when given the same seed. The CPU brand, network speed, even time of day all changed the results. On top of that, further inspection showed that the model wasn’t even implemented correctly in the software, leading to all the math being off.

      Without a professional statistician to review the model, and without a professional programmer to correctly implement it, the epidemiologist is likely to be spewing garbage… and doesn’t even know it.

      1. I don’t know a damned thing about epidemiology. However, I am a former professional statistician, and I can tell you with absolute certainty that majority of epidemiologists that are making predictions do not understand the statistics they are using.

        This is an absolutely stupid claim. This would be like someone who teaches legal history at an undergraduate school saying that appellate lawyers who practice at the Supreme Court don’t understand the cases they are reading.

        There is serious peer review in epidemiology. You can’t just get your BS published. Everyone’s statistics are carefully reviewed. They spend a lot of time constructing detailed models.

        Your training in statistics isn’t anything like what an epidemiologist does.

        And indeed, I think your comment (along with Nate Silver’s output) puts the problem in maximum relief- people who are statistically literate assume that they know everything about this subject, when in fact they are profoundly ignorant about it and would need years of training to get up to speed.

        1. “This is an absolutely stupid claim.”

          Sounds perfectly plausible to me.

          1. And the whole Monty Hall thing shows that many professional statisticians don’t understand statistics.

        2. No, this is like a chemist looking at what a patent attorney has written for their patent, and saying “No, this is wrong, this is wrong, this is wrong, this will never stand up because you’ve made errors here, here, here, and here”.

        3. Sorry, even an intoxicated undergraduate could see the flaws in the Health Nazi’s models…

        4. There is serious peer review in epidemiology.

          I think part of the problem here is that lawyers don’t really get the degree to which research in other fields is subject to review and criticism.

          It’s not just peer review by two or three others – not students, but actual researchers in the field. It’s also what comes before, and after, publication. There are presentations of the work, comments, etc., and plenty of people trying to prove how smart they are by finding flaws.

          IOW, obvious errors are going to be found, and found early.

          1. Are you familiar with what is termed the ‘Reproducibility Crisis’?

            What’s your take on the ‘In Medicine’ section of wiki’s article, and the papers it cites?

            (and separately, I haven’t looked at the model code myself, but I have read some of the code reviews … and it sounds pretty bad)

          2. This is quite optimistic… It’s not necessarily true however.

            Historically, Amgen took a look at 53 major cancer publications, and attempted to replicate the results. They couldn’t confirm the results in 47 of them…

            https://www.nature.com/news/biotech-giant-publishes-failures-to-confirm-high-profile-science-1.19269

          3. I’d think you’ll find that even peer review isn’t that great at catching errors. Scientists often push their peer review duties onto graduate students, or aren’t even experts at the specific work they’re asked to review. And after publication, there’s no reward for checking other people’s work. You can’t get that published, generally, so it sacrifices professional time to do something that doesn’t get rewarded.

            And model code is worse – it’s rarely published, and never peer reviewed. There’s no incentives to make code available. You can’t publish a paper on the implementation. So even your published methodology is useless if the code doesn’t actually implement it properly, but no one is reviewing that.

        5. You have no experience in statistics d you? Or in any analytical field, probably.

          I have worked with hundreds of academics in a wide variety of fields, and very few had anything more than an extremely basic understanding of what they were doing. The vast majority of “researchers” doing statistical analysis are just plugging numbers into pre-generated software packages.
          This should be expected – specialists in fields like epidemiology do not spend their time reviewing the strengths and weaknesses of different experimental designs, or under what circumstances you can do certain tests.

          As for peer review, ha! The example I specifically gave was the basis of dozens of ‘peer-reviewed’ papers by the creator, despite being flat out wrong. As Absaroka and Armchair Lawyer point out elsewhere, there is a major problem with ‘peer-reviewed’ research being impossible to replicate. Almost 1% of all ‘peer-reviewed’ papers published are later withdrawn. About 5% of all papers published are later discovered to have at least one significant error that did not lead to the paper being withdrawn.

          And despite your prejudice against statisticians, I can again explain: I don’t understand the details. I don’t need to. All I would do is look at the math. If the math is not right or if the assumptions and requirements for the model are not met, it doesn’t matter if you are discussing disease spread, chicken reproduction, or stellar dimming – your analysis is wrong.

          1. If the math is not right or if the assumptions and requirements for the model are not met, it doesn’t matter if you are discussing disease spread, chicken reproduction, or stellar dimming – your analysis is wrong.

            Well, Toranth, wrong mathematically anyway. But do you think it is worth remembering that when you model stuff mathematically, you reason by analogy? You posit numbers you suppose can be usefully treated as analogues for whatever real-world stuff you want to find out about. That strikes me as a bit different—meaning less clear-cut than, “if the assumptions and requirements for the model are not met.”

            Assumptions? Requirements? “Assumptions,” may leave room for imprecision, but “requirements?” Not so much. What is your call for “assumptions and requirements,” except a call for likely-unobtainable precision. And then, Analogies? Only the trivial ones can ever be precise. So there is some tension there, right?

            But back to the question of reasoning by analogy. Whatever the case with the uncertainties, the mathematical results you get back tell you how your real stuff would behave if your real stuff were numbers. And your real stuff is almost never numbers, or at least not entirely numbers (see again, “assumptions”).

            Once you notice that, you can imagine at least two different kinds of critique to apply to both the numbers and the other stuff. One kind of critique tests the internal consistency of the numbers. For doing that, a numbers expert ought to have a tall leg up on a researcher whose expertise lies elsewhere.

            But the other kind of critique strikes me as less of a clear-cut win for the numbers guy. The other critique is about how valid is the analogy these particular numbers represent. For certain (For certain, No. 1), your typical numbers guy is in over his head on that one. Except in trivial cases, the numbers guy has to borrow the researcher’s specialized expertise to even make a stab at evaluating the quality of the analogy—just as the other guy ought to rely on some expert’s numbers expertise. And, for certain (For certain, No. 2), the analogy is always imperfect. And not just imperfect with regard to numbers. Imperfect also with regard to ambiguities. There can be/will be contested ground whether the imperfections are modeling problems and fixable with better numbers, or whether the imperfections come from flaws or lacunae in the non-numerical expertise, or whether the imperfections come from calling on numbers to analogize stuff neither side understands well enough to permit doing that.

            That little list leaves for another time the question whether ambition to quantify every kind of experience must always founder on the rocks of Gödel’s incompleteness theorems. Do you have any thoughts on whether Gödel implies the possibility of expertise beyond the reach of numerical hegemony?

      2. Toranth, when you say you are a former professional statistician, are you telling us you used statistics professionally, or are you saying you were an academic scholar of statistical methods?

        1. I did statistical design and analysis for others; I have never been an academic researcher in statistics.

          1. Thanks.

    5. Dilan, I suggest that you read “COVID-19 pandemic modeling is fraught with uncertainties”
      https://physicstoday.scitation.org/doi/10.1063/PT.3.4493?utm_source=Physics+Today&utm_medium=email&utm_campaign=11551934_NQ+-+18-22+May&dm_i=1Y69%2C6VLJ2%2CJ2XN33%2CRLWP6%2C1&
      aNd then think again about how expert the “experts” really are.

    6. I don’t think Bernstein is claiming particular expertise here. He is making what is really a common sense observation, though as arch1 points out below, he is bit careless in saying,

      the author’s seem to assume that the victims were previously as healthy as their demographic average:

      since the study clearly took comorbidities into account, though not as accurately as was desirable, which the authors more or less admit in their response.

      1. By demographic average, I was including co-morbidities, but you are correct that I could have made that clearer.

        1. An 82 year old with heart disease who lives in a nursing home is not similarly-situated, life expectancy-wise, to an 82 year old who is otherwise doing well and is self-sufficient

          furthers the implication you weren’t including co-morbidities.

        2. David, you are writing for an audience which is keen to minimize and debunk the public health threat of Covid-19. Among that crowd, you ought to be careful how you handle imputations of cause and effect, because your audience will not be careful about it. And there is a seeming logical complication if you entangle co-morbidities with years of life lost, measured against life expectancy. I am suggesting there is a problem with your framing on this article.

          Note that co-morbidities which are themselves life-shortening do not have much chance to skew pandemic mortality statistics, because their life-shortening effects are already baked into life expectancy statistics. The life-expectancy at any age takes account of the usual incidence of mortality-affecting ailments. Whether it is Covid-19 which kills such a patient, or a prior life-shortening condition, not much statistical effect will register. In such cases, analysts using excess mortality as their gauge will not be misled into attributing to Covid-19 deaths of that sort, if Covid-19 did not in fact cause them.

          So if you do see a statistical effect of increased mortality during a pandemic, life-threatening co-morbidities are not likely important factors among the causes of the observed effect. Instead, you can be fairly confident that pandemic disease is the cause.

          Some of the excess fatalities the pandemic causes will indeed be patients with a different kind of co-morbidity—a kind which does not much dispose to an early death, except when it interacts with the disease (many auto-immune ailments work that way). In such cases, wouldn’t it be mistaken to posit a causal effect on the statistics to the co-morbidity, instead of to the disease?

          It seems to me that the minimizers are not taking care to disentangle the means of measurement, from specific co-morbidities, from the actual causes of the measured deaths. Probably, public health officials are also confusing the public if they attribute deaths to co-morbidities without first distinguishing the contrasting causal impacts among them.

          An especially potent source of confusions of that kind are the public announcements saying that old age is a complicating factor leading to increased risk from Covid-19. At the level of helpful advice about managing personal risks, those announcements make sense. But aren’t folks mistaken who infer from that a conclusion that old age is a cause of Covid-19 deaths? Old people are Covid-19’s principal victims, but their age does not cause their deaths from the disease, not even when measured statistically—the disease causes that.

          1. David, you are writing for an audience which is keen to minimize and debunk the public health threat of Covid-19.

            I’d say that your referenced audience is keen to draw a distinction between the old and sick and rest of the population. If, as it seems increasingly likely, the case fatality rate among the younger and healthier cohort of the population is in the low .x% range, or possibly even somewhere in the .ox% range, then this should drive public policy. This level of risk should in no way preclude normal activities among this group. For the old and sick of course a much higher level of risk avoidance makes sense.

            1. The audience is grasping at whatever straw they can, whether that’s overcounting or co-morbitities or speculation about economic causes of death or just bare yelling about tyranny.

              But even here, it’s actually not that many people, just a few very loud ones.
              I’m gratified at many of those here who I often dismiss as being terminally outcome-oriented that do seem to be taking this seriously.

          2. I’ve been wondering about the emphasis on quality years of life recently, as opposed to mere deaths. While an interesting statistic, in this political time, it must be planned for use in attacking, or defending, Donald Trump. I just haven’t figured out which, yet.

    7. So Dilan,

      You always believe the experts in the field, without reservation, no matter the field?

      Or do you make judgements about their comments and the fields?

  9. There is an additional piece of data that may share some light on this if available. That is, the demographic if those that die with a do not resuscitate order. The larger this fraction of the age cohort the smaller effect on average years of life lost. DNR orders are not uncommon in managed care facilities.

  10. This wasn’t super clear in the article so I thought it should be mentioned.

    The study in question didn’t assume that an 82 year old who died of COVID-19 would on average, live as long as a healthy 82 year old.

    It assumed they’d live as long as an 82 year old with the same documented co-morbidities (heart disease, diabetes, etc).

    The issue is how to account for the fact that the victims are likely to have had other undocumented co-morbidities.

    It’s a very valid point, though this is a point that epidemiologists are well aware of and I suspect they have some idea of the magnitude of the effect.

    1. That’s one issue, the other issue is that saying someone who is 82 with heart disease and died with Covid was likely of average health of someone who is *82 with heart disease.* In fact, some 82 year olds with heart disease are much sicker than average, and one might predict that they will be more likely to succumb to Covid than average. The authors rejoin that they have no way of measuring that, so they don’t. They also point out that maybe sicker people are self-isolating more, and that could be a confounding factor.

      1. Sicker people might self-isolate more, IF they have the choice. People in nursing homes don’t have a choice about whether COVID-19 patients live in their facility – and in at least two of the states with the worst COVID-19 death rates (NY and Michigan), the state government _pushed_ COVID-19 patients into nursing homes.

  11. Prof B You simply can’t compare an otherwise healthy 82 year old with heart disease to someone whose heart disease so enfeebles him or her that they need to be in a nursing home.

    Study guy I think JB and DB may be making a different statement that “notwithstanding the fact that the average life expectancy is still quite long among older people with comorbidity, those dying from COVID-19 are likely atypical compared to the average among older people with comorbidity”.

    That looks like the same question

    Study guy : I think we are talking here about residual confounding, i.e., after you take into account the known/measured variables, are there remaining differences between patients on which we estimated life expectancy (the general community in Wales) and those dying of COVID-19 in the Italian data.

    That looks like a different question.

    DB’s question seems to imply that comorbidities (eg heart disease) may come in different doses. Someone with warp factor 10 heart disease is more likely to be carried off by COVID than someone with warp factor 3 heart disease. Thus “heart disease 10” and “heart disease 3” should be treated as different comorbidities.

    So what we’re looking for is not some secret sauce, other than known comorbidities, that distinguishes COVID fatalities from the general population, but more precision in the specification of the comorbidities that have already been identified.

    1. “So what we’re looking for is not some secret sauce, other than known comorbidities, that distinguishes COVID fatalities from the general population, but more precision in the specification of the comorbidities that have already been identified.”

      There are certainly important moving parts beyond this, Lee. E.g. as David B pointed out, there can be unknown comorbidities, which may correlate with other data which *are* available (such as level of preinfection assistive care).

  12. David, your comment on the paper made 3 good points, all of which were acknowledged in one way or the other by the lead author (whose response also added a useful observation about possible selection pressures in the opposite direction) – good job! IANAS, but it’s surprising to me that potential residual confounding effects were apparently not acknowledged more directly in the paper.

    Nit: I suggest that in the OP intro you say “…demographic *(including known comorbidities)*…”. As it is, the first clue I had that these were already accounted for in the paper was the final sentence of your comment (I’d taken “coded with the same disease” in the first sentence as a reference to COVID itself).

  13. I think the notion of estimating an average of lost lives is interesting, but I don’t think we have the required microlevel data to make a decent and useful estimate. The reason is that one can think of two separate averages. One would be the loss of expected life-years with no intervention, i.e., letting the disease run its course unimpeded, the other the loss given some intervention to mitigate the spread. Obviously, to understand the effect of the interventions (social distancing, masks, medications, economic shutdowns, etc.) we should like to compare no-intervention data with intervention data in appropriate statistical samples. We don’t have that, and will never get it, because interventions started immediately. All we will ever have are data on morbidity and mortality given existing interventions. The problem with the average we have is then that the interventions are not uniform across the data — they add a level of heterogeneity to the data that we need to disaggregate to analyze. For example, if the data include downstate NY, the forced admissions of positive patients to senior living facilities will skew the data very much compared to the rest of the country. So, for the US, we need at least to disaggregate that area from the rest. But the rest of the interventions are also heterogeneous across states. That is, there is no single set of interventions that would answer the question of what loss of life-years has been experienced — it depends on a long list of various interventions across areas and doctors. All we get will be some number that does not tell us what to do next. I guess I find that of some interest, but not very helpful. We already know that very young people die infrequently, and old age folks with co-morbidities die at a much higher rate than others. This is relevant, but averages not so much. I think.

  14. Like many, I don’t know anyone who’s died of covid. Have a son who was told he had it- after 2 negative tests, so I may or may not know anyone who’s had it. Covid is the Schrödinger’s Cat of diseases.

    Taking with a friend today- in his driveway. We were less than 6 feet away from each other sans masks. Living dangerously. Anyway, he knows one of the 2 people in out county whose deaths have been designated as covid deaths. He was in hospice, pretty much waiting his turn to die. Covid may have shortened his life span by a day, maybe two whole days. Or may actually have had nothing to do with his death, just something he had when he died. No way of knowing. So I’ll say it again- Covid is the Schrödinger’s Cat of diseases. It may or may not be responsible for a particular death. Even if it is observed.

  15. “How Many Years of Life Does the Average Covid-19 Victim Lose?”

    All of them, for the most part.

    Which is why this sort of analysis is problematic with deadly (as opposed to just fatal) diseases. Would you, given two strangers, and a magical elixir on a trolley, take all of one stranger’s two remaining years, or five years from an apparently fit and hale fellow?

    Mr. D.

  16. “Studies” like this masquerading as legitimate science are dangerous. There is no way based upon the extremely limited evidence we have from Covid that one can predict, with any meaningful success, the long term impact even beyond six months let alone proclaim it takes ten years of life. This is just irresponsible to publish in the first place.

  17. I’m not a scientist or a mathematician or a computer-ist. I’m a lawyer (now retired). But I dealt with plenty of experts of various kinds in my practice. When I took the deposition of an opposing expert, one of the important questions was: What was the purpose of your study? (This was also an important question in working with the expert(s) on my client’s side). If you are studying “excess” deaths due to coronavirus in assisted living facilities for the purpose of advising a life (or health) insurance carrier regarding adjusting their rates, you’ll properly go about it differently than if you’re advising the estate of a decedant regarding a wrongful death claim. Both ways may be valid for the purposes for which they are designed, but not for a purpose for which they were NOT designed.

    So, why do you want to know how many years of life this victim/these victims lost?

    1. So true. Reminds me of what I instructed my finance people at my company to say, and it applies to any company, when asked “What does it cost to make product X?” Respond with: Why do you want to know? Unless you know how the answer will be used, you can’t answer the question.

  18. “one never fully recovers from things like pneumonia” Any medical authority to support that incredibly sweeping claim? Or you just intuit it?

    “[b]fascist shutdowns{/b}.” Are you really serious about “fascist,” or is that just an excessive expression of your disagreement with the quarantine and other public health measures that have been adopted irregularly across the country to control the covid-19 contagion here? If serious, please explain how you imagine “fascist” an apt political descriptor. Do you think these same sorts of measures have been “fascist” elsewhere, like Italy? Would you agree that the Peoples Repubic of China, not generally thought of as a “fascist” regime, though inclined to totalitarianism, employed more draconian measures than we have here in the US?

    “Suicide is up, domestic violence is up, drug abuse is up, while treatment for the top two killers (Cardiac & Cancer) is way down.” Any reliable data to support that contention? (The Federalist piece’s collection of unverified testimony from some practitioners
    hardly qualifies as even anecdotal evidence.) These “projections,” surmises, or whatever they are, may have intuitive appeal to you, Trump, and other politically like-minded thinkers, but they have little or no probative value. https://abcnews.go.com/Politics/fact-checking-trumps-claim-suicide-thousands-economic-shutdown/story?id=69790273

    Do you think Fauci and the other prominent public health officials who have advised these measures lean to fascism? (Maybe Pat Buchanan, if he’s still alive, will come forward to opine on fascist methods for disease control.) Are there any public health authorities here or elsewhere that you have some confidence in and think we should be listening to?

  19. How do you imagine covid-19 would reduce the chances of another fatal illness or postpone one? I can’t.

  20. Thanks admin for giving such valuable information through your article . Your article is much more similar to https://www.bocsci.com/product/coa-s-acetyl-tryptamine-8889.html word unscramble tool because it also provides a lot of knowledge of vocabulary new words with its meanings.

Please to post comments

Comments are closed.