Coronavirus Grab Bag: deaths vs qalys, safety vs safety theater, ‘all in this together’, and more.

This post is by Phil Price, not Andrew.

This blog’s readership has a very nice wind-em-up-and-watch-them-go quality that I genuinely appreciate: a thought-provoking topic provokes some actual thoughts. So here are a few things I’ve been thinking about, without necessarily coming to firm conclusions. Help me think about some of these. This post is rather long so I’m putting most of it below the fold.

I. Deaths vs ‘life-years’ lost, with or without ‘quality adjustment’.
I have pretty much been looking at ‘deaths’ as a key metric for the severity of the pandemic, but I’m not happy with it. A big issue is that all deaths are not equal: on average, the older the victim the smaller the tragedy. Here’s how I think of it: I have about 30 years of expected life ahead of me. A 20-year-old has about 60 years of expected life ahead of them. If each day of my life is just as precious to me and to society as each day in the life of a 20-year-old, if I die the net loss is half as many of those precious days as if the 20-year-old dies. Certainly I’d have a hard time arguing that each of my remaining days is _twice_ as valuable as that of a randomly selected 20-year-old. And how about a 90-year-old with dementia who has just been diagnosed with pancreatic cancer? If there’s a fire, and a fireman has a choice between saving this person vs a 12-year-old, would anyone seriously argue that it should be a toss-up? I hope not.

Rather than ‘deaths’, or in addition to it, I’d like to see ‘life-years lost’. When someone dies, you can use actuarial tables to determine how many expected years of life have been lost. Sum that over everyone who has died of covid-19 and you’ve got a number I would rather see than the raw count of deaths.

There are various options, like whether to adjust for the ‘quality’ of the years lost (that 90-year-old with dementia being assumed to get less out of each day than a healthy 20-year-old)…my personal preference would be to _not_ have someone do the quality-adjustment for me, I’d rather have just life-years. But QUALYs would be much better than just ‘deaths’.

Why isn’t it happening? Is there any news outlets that routinely reports on life-years lost?

On a related subject, I wonder: to what extent are differences in opinion about the pandemic response related to differences of opinion about either the importance of life-years lost or the appropriate ‘quality adjustment’ for an old person’s life? To come back to Sweden, which I have discussed in an earlier post, the Swedes seem very happy with their choice of doing mostly just voluntary measures to control the pandemic, in spite of the fact that their death rate is among the highest in the world. As with every country, most of the people dying are quite old; they’ve had about 3200 covid-19 deaths, which is 320 deaths per million population; the U.S. is officially 240 per million, although we surely have a bigger undercount here than they do. Still, they’re probably 30% higher than us, but they don’t seem bothered. Should they be? It’s not like a randomly selected 320 people per million have died in Sweden: nearly 2/3 of the deaths have been people 80 years old or older, and only about 5% have been under 60.

II. Safety vs Safety Theater.
This is on everyone’s mind one way or another. There are rules that make sense to me and rules that don’t, and I wonder whether the ones that don’t are just supposed to make us _feel_ like they’re doing something, or whether they serve a purpose I don’t know about?

For instance, here in California you are allowed to take a walk in the park with members of your household. But you are not allowed to have a picnic in the park. A walk in the park is said to have an important health purpose — gotta get your exercise! — so it’s allowed in spite of whatever virus transmission risk is associated with it. A picnic is said to serve no vital purpose and thus to be unjustified. This doesn’t make much sense to me.

I accept that either activity does involve some risk of viral transmission. Most people who drive to the park will do so in a car that needs gasoline. All those people driving to the park means more stops at gas stations, gotta touch the pump, maybe someone else is nearby, etc. If the park has other people then you’ll walk past some of them and inhale or exhale some air they exhaled or inhaled.

And if you were to picnic at an actual picnic table, well, you’re touching the table and the people who come after you will touch the same table, so there’s that. Also, if people are picnicking then they’ll generate trash, which means you need trash pickup, and that trash has to be dropped off at a transfer station, and all these things — having people driving garbage trucks and manning transfer stations etc — all increase the number of people exposed to other people.

So, OK, have a million people go on picnics in the parks over the period of a month and the number of additional transmissions won’t literally be zero. But c’mon, how many additional transmissions would happen this way compared to the number that happen anyway from people going about permissible activities like shopping for groceries? And to the extent that having a picnic in the park with your family is good for one’s mental health, especially as an alternative to being cooped up in a small apartment together, does it really make sense to deprive people of this minor pleasure?

I don’t think picnicking in the park with one’s family should be illegal, or even discouraged. But: I could be wrong. Seeing people picnicking in the park as life is normal, maybe that gives people a false sense of security and would nudge people towards dropping their guard on more consequential behaviors? Or maybe the governor or his advisors think the risk of transmission to another household is higher than I think it is; for instance, maybe they think children from different families will inevitably intermix because in practice their parents won’t be attentive enough to stop them.

This is just one example of something that I think isn’t actually significant when it comes to slowing the spread of the virus, but that with which we are expected to comply. I am not a scofflaw on this stuff: I think it’s safety theater and doesn’t actually improve safety, but I don’t know for sure and I’m not going to pretend I do. I just hope these decisions are being made by someone who is paying attention to reality of the risks rather than the ‘optics.’

III. Are we all in this together?

Boy, am I lucky: so far my work is unaffected by the pandemic. My annual income in 2020 will be pretty close to what it was in 2019. My wife’s might be down a bit, but not a lot. We are doing absolutely fine. But some of our friends and neighbors are (already) out of work, and our friends who are out of work are in a much better financial position than many or most people in the country who have lost their jobs. 

I’m sure my circumstances influence my view of the response, in the direction of making me want to ‘err on the side of caution’ when it comes to relaxing government regulations. I see news articles about people complaining about the enormous sacrifices being required, how much it’s crushing the economy, etc. etc., and I think “jeez, if you think this is ‘sacrifice’ what do you think about World War II, when millions of people were drafted and sent off to fight, while back home there was rationing of meat and gasoline and rubber?”  A friend mentioned how unreasonable it is to expect people to go for months without being able to hang out with friends or see their girlfriend, and I pointed out that the men storming the beaches at Normandy also had to go for a long stretch without seeing friends or girlfriends, and that many of them never got to do those things again at all. As sacrifices go, what is being asked of us now is huge, but in the past people have been asked for much, much more.  To which a perfectly appropriate response is: “Sure, Phil, easy for you to say.” And it is. I’m not suffering. 

It isn’t fair. I’m not a big fan of the sentiment that ‘life is unfair; get over it’. The world would be a better place if it were a bit less unfair. Maybe a lot less unfair. What are some politically feasible ideas for spreading the pain? Anyone got any ideas? 

 

This post is by Phil.

181 thoughts on “Coronavirus Grab Bag: deaths vs qalys, safety vs safety theater, ‘all in this together’, and more.

  1. Phil:

    In answer to your first question: Yes, I too like the idea of counting life-years lost. But it’s tougher to count life-years than to count deaths. If 100 people die, that’s 100 deaths, period. How many life-years were lost? You can only estimate. I agree that we should be estimating it. But I can see why deaths will be reported, as these are direct numbers. Consider that they continue to report the number of positive tests: it’s really not clear what to with that number, but it’s a number, so it gets reported.

    • Andrea,

      This was a great read. Thank you for sharing!

      What I found most interesting:
      – SMR — e.g., “the standardized mortality ratio (SMR) summarizes how a given comorbidity can increase the risk of dying” — has a dramatic effect on “life-years lost” (no surprise; but it is nice to see the math and background well-presented!). For example if most people who die from COVID-19 have SMR=3, then they would have died in the next ~2-4 years. If SMR>4, then total “life-years lost” quickly approaches 1 year or less.
      – If the vast majority of hospitalizations and deaths from COVID-19 have SMR>=2, then we may want to re-evaluate our view of the risk it poses to the general population (by this I mean those having SMR=2 is not unreasonable to assume — which would mean those hospitalized/dead from with COVID-19 would have done so within the next 2-4 years.

      • I really think that as we begin the whole “reopen” thing the demographics of who get infected will shift to a younger group, and many more 40-60 year old people who’ve been sheltering at home will get this thing and die. The fact that the initial burst of people affected were the most vulnerable, in LTCFs and soforth means that “so far” we have one outcome… but soon we’ll have another outcome. Every 50 year old who dies takes a toll in life years 5-20 times as big as an elderly or LTCF person. My guess is that the peak rate of life years lost is still coming. we may well be past a peak in deaths but the real life year toll will come by mid summer or early fall as more and more people are exposed, because we’re doing *zero* absolutely *zero* case control type work.

      • Daniel,

        Maybe?

        Your conclusion depends heavily on knowing not just age, but the comorbidities of those who die — e.g., the SMR score from the resource Andrea shared.

        Someone with SMR=1 who dies from COVID-19, as you state, will cost 20x life-years-lost compared to someone in a LTCF. Someone with SMR>=3 at age 50, however, likely has much lower life-expectancy (8 years or less). Hence why reporting comorbidities at death is so important.

        So if we re-open and many people with SMR~=1 (e.g., relatively healthy) start dying, the I agree — we have a major problem. If we re-open and only people with SMR>=3 start dying, then we still have a problem — but I’m not sure we will see a massive increase in life-years-lost.

  2. Phil-

    Life years are a valid thought…whether we characterize it as such, we make medical decisions for ourselves or others with this rough metric. One can estimate it fairly easily oneself by weighting the death numbers with their age distribution of course, but that only solves the issue for you individually and fails to be discussed by media reps of either polar stripe. I would put the reporting of raw case numbers rather than percentage positive or other weighted metrics as another source of frustration in tracking the disease and it’s effects.

    Len

  3. Phil –

    > I have about 30 years of expected life ahead of me. A 20-year-old has about 60 years of expected life ahead of them

    I’m going to assume that you’re younger than I am.

    Is your life worth more than mine as a result, because theoretically I will die before you?

    And you’re only picking one of many relevant metrics. How about my health vs. yours. Should we adjust the value of one person’s life relative to another based on an actuarial table and their health history? How about how much we exercise? And then into health history, you’d have to factor in SES, since it has a strong effect on health outcomes. Should we value married people more than unmarried people, since married people tend to live longer? Women more than men? Whites more than blacks?

    How about how many people depend on me compared to how many people depend on you? How about where we live? How about our occupation?

    • I’m not exactly dismissing the issue. I”m a big believer in allowing people to factor in age and/or health condition into making decisions about end of life care, for example. But I also think that this is a very complicated issue. IOW,

      > When someone dies, you can use actuarial tables to determine how many expected years of life have been lost.

      I think that you’re going to have some pretty complex actuarial tables there even if you just wanted to base them on “expected years of life lost,” but even in just singling out that one metric you’re making a value judgment that just glides by a lot of pretty subjective issues.

      Let me ask you this. What would you hope to do with this evaluation of “expected years of life left?” What insight do you expect to gain if news outlets report on this metric you want to see used?

    • The right frame of mind here is what’s called an asymptotic series. You make first order corrections, and then ask if the answer is good enough or you need second order corrections. The corrections between a 13 year old dying and a 98 year old are ENORMOUS (let’s say the 98 year old has 0.5 years to live and the 13 year old has 80, so it’s a factor of 160), the corrections between a white 13 year old and a hispanic 13 year old are TINY by comparison (maybe a factor of 1.15 ?)

      We should absolutely be doing the life years lost as a function of age correction, because it’s an enormous difference. The rest I think we can let fall by the wayside.

      • Daniel –

        > The corrections between a 13 year old dying and a 98 year old are ENORMOUS

        So then what are your cutoffs? If we say we should consider the 13 year old’s life more valuable than the 98 year old, what about the 40 vs. the 50? Seems to me that ultimately you’re going to be determining some cutoff point, in some kind of arbitrary fashion. Or maybe not. If not, what is the rationale?

        And how do you determine which metrics have a large enough significance? I’d say that race/ethnicity have a pretty big signal in life expectancy. For obvious reasons, I’m sure you’d be reluctant to use that is a metric. How do you compare cut offs. If race/ethnicity has bigger signal across categories than 10 years does within a given category (just as an potential example), how do you determine to use the one metric but not the other?

        And again, I’m asking to what purpose to you expect to use this information. I don’t know. I think that sometimes people want to measure things because they think that measuring things gives them some kind of useful information, without really determining if that information is fit for purpose. What is the purpose of calculating expected years lost?

        • I see. So just a total. Ok. Are you going to use this information as the basis of some kind of decision?

          But why would you just add up the ages but not other metrics? You’re going to calculate age lost for a 40 year old who dies And a 41 year old who dies. But then why not factor in other variables that have a significant impact on life expectancy? Let’s say comorbidities as an obvious choice. What would you get from counting the 1 year less from the 40 year old compared to the 41 year old if you don’t also consider the 39 year old that has end stage cancer?

          I’m really not trying to be obtuse, but I don’t get the purpose – other than to just get a number that somehow seems like it means something.

        • I just want sum(life_ex[i] – age[i]) for i across all the people who died, because I want to know exactly how many life years this virus is erasing. And I’m willing to settle for an estimate of life_ex[i] which depends only on the age[i] and not a bunch of other factors, because I am willing to put up with a certain amount of uncertainty and don’t feel like it’s very important to minimize it to its smallest possible value.

          You might as well ask “why count deaths? who cares?”

        • Daniel –

          > You might as well ask “why count deaths? who cares?”

          Well, I actually thought about that. I mean a lot of people are arguing that the number of dead is pretty irrelevant in a sense. The argument seems to go that “after all everyone’s going to die. And maybe what really matters is our ‘freedoms’ that are being stolen from us by librul petty tyrants, just so they can help some old people live another 6 months or so.”

          But I think there’s a difference, because of the binary nature of dead vs. not dead. But yeah, there some subjectivity there also.

          I still don’t think I’ve gotten an answer to my question of how this information will be useful to you. Seems to me that you’ve basically just said that you want it because you want it. I guess maybe I’m a bit uneasy about the implications of going down that road unless there’s an actual answer for why the information is useful. But sure, there’s no reason on principle why I’d object to anyone obtaining that information. And if you think it’s useful then I’m not really in a position to judge that.

          > because I want to know exactly how many life years this virus is erasing.

          But you’re not really getting that because you aren’t factoring in a lot of the relevant metrics.

          > and don’t feel like it’s very important to minimize it to its smallest possible value.

          But the age at death, particularly unless it’s a dramatic contrast, is often considerably less of a small detail as compared to something like whether someone has end stage cancer, or even race (there’s like a 12 year difference in life expectancy based on race between Asian Americans and African Americans).

          OK. I give up now. I’m really NOT trying to piss you off.

        • Joshua. sure I think if you’re trying to estimate the life years lost *for a particular person* things like cancer status, and atherosclerosis, and whether they drink heavily or use drugs and soforth are all really important factors. But we don’t have detailed health status on each person, that’s private information, so we’re inevitably led to estimating based on the information we are likely to get, which is mostly their age. It doesn’t make the inference inaccurate, just more uncertain.

        • > It doesn’t make the inference inaccurate, just more uncertain.

          I don’t know what’s your definition of “inaccurate” but conditional on age there will be a correlation between life expectation and probability of dying of COVID19 if infected. Your calculation won’t tell you exactly how many life years have been lost on the particular people who died, it may tell you exactly something else.

        • Carlos, it’s true without knowledge of the comorbidities and soforth, we’ll have to accept larger error bars, some of which will be down to unknown biases, which we may have priors about, so we can get a bunch of different posterior distributions based on what our priors are about the biases inherent in estimating life expectancy without sufficient information… but none of that should be too shocking to this audience right?

    • Joshua,
      As you’ve both implied and suggested, age alone does not tell us, or at least doesn’t tell me, what someone’s life is ‘worth’. There are young serial killers whose life is worth much, much less than some 85-year-olds, to me and I think to a lot of people. Consider that, as a society, we think some people’s lives are worth less than zero, so we kill them.

      When it comes to making public health decisions, we make them wholesale, not retail. Any individual young person’s life might or might not be worth less than any individual old person’s life, but we should definitely consider 10000 randomly-selected young people to be ‘worth’ more than 10000 randomly-selected old people. Otherwise we’re saying that a day of a young person’s life is worth much much less, on average, than a day of an old person’s life, and I don’t see any defensible argument for that.

      As for how detailed the life expectancy tables should be, I dunno, it’s worth discussing and I don’t have a firm answer. I definitely think we should look at age. Once you’ve done that you’re 80% of the way to doing everything anyone could ask you to do, so let’s start with that and then discuss whether it’s worth going farther.

      And of course, there’s no reason things have to be summed up in a single number. I’m not happy with just the number of deaths. I’d be happier with number of deaths _and_ number of life-years lost based on age alone. If someone gives me those and then _also_ offers me the number based on (age + health status) or something, that’s even better.

      • Phil,

        > As for how detailed the life expectancy tables should be, I dunno, it’s worth discussing and I don’t have a firm answer. I definitely think we should look at age. Once you’ve done that you’re 80% of the way to doing everything anyone could ask you to do, so let’s start with that and then discuss whether it’s worth going farther.

        The resource Andrea Manca shares above explains this well: https://avalonecon.com/estimating-qaly-losses-associated-with-deaths-in-hospital-covid-19/

        We must consider comorbidities in addition to age, as comorbidity has a strong effect on life-years-lost and IFR that is independent (but does correlate) with age. Especially since the majority of deaths in Italy, NYC, etc., seem to have the majority of deaths having comorbidities.

  4. We pretty much agree that life-years is better than deaths for quantifying this tragedy. As proof, pretty much everyone has a different reaction to the death of a 23-year old versus a 93-year old.

    But, we have no intuitive feel for life-years, so it doesn’t give us good intuition about the magnitude of the tragedy. I understand how many people 100 people is. I have no feel for 3,000 life-years though. Even millions of deaths I can understand as a percentage of the country’s population or the number of deaths from other causes annually.

    So, we need something that combines both. Something like “x deaths with a median age of y,” or “x deaths of people 60 and older, and y deaths of people younger than 60.” And this is exactly what is commonly done; because it makes sense. Indeed, since it is widely known that most deaths are of older people, we automatically make the adjustment when we hear simple death counts reported.

      • That’s another way to do it. Takes a little effort to grasp it, but it is graspable and conveys useful information. For general consumption, though, it isn’t optimal because you have to educate the reader, and in most instances, you can’t stop to do that.

    • I don’t agree with that. I have much different reaction to my grandfather dying at 93 than to a stranger dying at 23. I also think that, on average, individuals value their own lifes more than they value others lifes.

      Here is a morbid thought experiment:

      Take 100 randomly sampled 93 years olds and pair them with 100 randomly sampled 23 year olds. Put them in sealed chambers where each individual in a pair can see the other, but not communicate. Give the 93 year old person a switch and tell them “Only one of you will leave here alive. Turn to the right and you will walk. Turn to the left and the other will walk.” How many 93 years olds will turn the switch to the left? Do we think it would be an significantly different number than if the 23 year olds were given the switch and the same command?

    • You can get life tables from mortality.org as CSV files. You have to “join” to get access but it doesn’t cost anything–just accepting their policies about how they are used (which should not be any kind of barrier to anything being discussed her, nor for research in general.)

  5. 1.
    In theory, life-years might be a better measure than deaths, but the problem is that it opens up a whole can of worms.
    While most people might agree that saving a 20-year-old should take precedence over a 90-year-old, sizable number of them will switch once the alternatives are between saving a 20-year-old or two 90-year-olds.
    Furthermore, I doubt you will get the majority to agree that it is better to save a 20-year-old than to save ten 90-year-olds.
    Of course, once you start quantifying a person’s rest of life, the natural next step is to assign weights depending on factors such as individual happiness, social contribution, etc.
    And that is a field full of landmine.

    If you are really interested in this issue, I suggest checking out ‘Weighing Lives’ by John Broome.
    It is an extremely complicated problem with no easy answers.

    • The decision theoretic issues are irrelevant to the measurement issues. “How many life years did we lose?” is a straightforward estimation question. “How much do we value XYZ combination of life years?” is a totally separate question.

      • > The decision theoretic issues are irrelevant to the measurement issues.

        Yes, but why bother measuring something if it is not going to affect anybody’s decision.
        We already know most of deaths are concentrated among the elderly.
        What would measuring life-years tell us that we already don’t know?
        Or to put in other words:
        How should we react differently if the measurement of life-years turns out to be smaller/greater than we anticipated?
        Should we react differently?
        If not, again, why bother?

        We measure death partly because it is something we want to minimize ceteris paribus.
        Can we say the same for life-years?

        I suspect what Phil (and possibly you) has in mind is something like
        ‘instead of minimizing deaths x, we should minimize f(x, y) where y is life-years and f(x,y) is increasing in both x and y’

        I think this is something a lot of people can get behind and I don’t disagree.
        But here are the problems.

        1. Measuring y is difficult and comes with huge error bars. So many countries are struggling for a good estimate of x, a much easier task.

        2. Once you start measuring life-years, you should really look at more than deaths. Just because somebody survived Coronavirus doesn’t mean there was no loss in life-years. Good luck measuring that.

        3. Even after you get a decent measurement of y, there is no point of reference for people. So what if you conclude that 60,000 lifetimes were lost due to coronavirus? What does that mean? How many lifetimes were lost in 9/11 or Vietnam war? This becomes a bigger problem combined with 2.

        4. It will take a long time for the society to come to some form of consensus on what f(x,y) should look like.

        • Your argument comes down to something like “what you don’t know won’t hurt you” and I think it’s an odd attitude at a statistics blog, but sure I can see that you’re already far down some political trail trying to put up roadblocks to a thing you dislike the idea of. For me, I’d just like to know how bad has this thing been, and is it accelerating or decelerating in terms of life years lost. My basic value set is that the more information I have about what’s going on, the better.

        • I believe my argument is closer to “what you don’t know, but think you do, will hurt you more than what you know that you don’t know.”
          To be clear, I’m not against the concept of life-years, although I understand why you might think so.
          I personally do think there are many different forms of f(x,y) as a measure of social suffering preferable to x.
          But people should be aware of what they are getting themselves into.

          Last food for thought:
          How did Coronavirus affect the life-years of the population that were not infected?
          In a way, this is both argument for/against life-years.
          On one hand, it shows that deaths is lacking as a measurement for social cost and motivates development of additional measures.
          On the other hand, it again shows the difficulty of measuring life-years with reasonable accuracy.

        • You can measure anything. “How many convicted rapists have been killed by Covid 19” is also a straightforward estimation question. But before anyone goes to the trouble of compiling it for you, you’re going to have to explain why it’s useful to you.

          It’s false of you to suggest there’s not going to be some value judgment inherent in what you do with the stats. And thus it’s totally reasonable to discuss those value judgments and what we’d be able to accomplish with the facts before collecting them.

        • Fred,
          I think our response (by ‘our’ I mean our government, our society, our culture) should be different for a disease that kills mostly old people compared to one that kills mostly young people. I feel the need to say here that this doesn’t mean life-years list is the only important metric, that deaths is the only thing that matters, or any other such simplification. I’m just saying the age of the victims should matter, if it is radically different from a random cross-section of society.

          I understand slippery slope arguments but I am skeptical of them. Deciding that life-years matters is not the same as saying that it’s all we will pay attention to. I don’t think adding this metric to the reporting will set us sliding down a path that leads to putting grandpa out on an ice floe. I don’t think I or anyone else needs to explain exactly what we will do with that metric, any more than anyone has to explain why they want a count of deaths, or cases, or critical cases, or hospitalizations. They are all metrics that help characterize the impact of the pandemic.

  6. My response to people who say that someone should not see a boyfriend/girlfriend: If they are married, they should social distance from their spouse for the time period. This is not riskier than people who live alone seeing each other, as long as they are not promiscuous.

    I dislike the approach of using one person’s suffering to diminish another’s.

    Anyhow—adults can still see their girlfriend/boyfriend: They can meet for hikes and in most places nobody is going to stop them from going to one person’s home together. In the event someone does inquire, you can say you are helping them or bringing them food. In fact one local newspaper agreed it was fine. to visit a long term partner.

    • Yes. My girlfriend and I were traveling carefully together just as the pandemic emerged in our area, then we stayed apart for 7 days (beyond the average incubation time) to be quite sure we weren’t sick, and since then have been seeing each other about as usual—even crossing the DC/Maryland line to do so, against the rules of her state. We violate the letter of the stay home orders but it seems much safer than other things we could do within the rules. At the same time, we recognize in principle that people shouldn’t second guess the public health edicts … much?

  7. I’m not sure life-years lost is really a metric that deserves attention, even assuming you could measure it accurately. Every infection is equally bad in terms of taking up hospital resources and “flattening the curve”…. a 90-year-old in the hospital and a 40-year-old in the hospital both use one bed after all.

    • But they may (probably will) use more hospital resources than just beds, and the different resources they use mayt have different “costs” in the sense of interfering with treatment for other patients.

    • Every infection is decidedly not equally bad. Some infected never become ill, some only mildly so. Only the worst cases need treatment, particularly ventilation, and the chance of that happening (conditioned on being infected) is very dependent on the age of the patient. Also, this has nothing to do with reported deaths, if we care about hospital admissions then we should report hospital admissions, not fatality numbers.

    • Taking up hospital resources is a relevant metric only in places where there is a realistic risk of overwhelming hospital resources. This does not seem to be the case in much of the US. Even with “reopening”, people aren’t going to totally go back to 2019 behavior, so we aren’t going to see totally unrestrained exponential growth (like 3-4 day doubling times).

      In my state (TX), at least, it would take something like that to overwhelm the hospitals, given a lag of three or four weeks before the effect was obvious and measures could be re-instituted (either formally by the governor, or informally by people reacting to the surge of hospitalizations and/or deaths). Un-used capacity is more than 10x the capacity used by COVID patients.

  8. Lakeland: “The decision theoretic issues are irrelevant to the measurement issues”. Bloom: No! One’s decisions and one’s latent commitments drive what one chooses to measure.

    • you misunderstand, if you want to calculate the ‘Utility’ or ‘value’ of the lives lost, you need to have a function U() so you can do

      sum(U(life_years_lost[i]))

      but if you just want to calculate the expected years lost, you can sum them up and the ‘value’ doesn’t have to *enter into the calculation*, you can just do sum(life_years_lost[i]) under the best available actuary model.

      I’m not arguing against the idea that it’s a value judgement to care about this statistic, sure it is, but the values of different people don’t need to enter into this count directly.

  9. UBI is the obvious practical answer to III. Is a global pandemic enough to make it politically tenable in the US? Sadly I doubt it. We took a step in that direction… a one-time check for nearly everyone! Proudly signed by the POTUS! But the resentment to the possibility of another round is palpable.

      • “Well people in hell want ice water too,” said Sen. John Kennedy (R-La.), asked about another round of checks. “I mean, everybody has an idea and a bill, usually to spend more money. It’s like a Labor Day mattress sale around here.”

        • So, you have some Republicans who do not care about being elected in 2020 making these arguments and you are characterizing this as “resentment to the possibility of another round is palpable”?

          As a Democrat, I find this to be a huge political opportunity. I would plaster that quote in every ad I could between now and November. But, as a citizen, I find it lacking in both intelligence and understanding of what is actually happening right now.

  10. I agree that life-years lost would be a more valuable metric. But how to measure? You need to use conditional survival estimates; conditional on whatever underlying health issues the patient had to avoid over estimating. That would be quite a bit of work.

    I have, however, seen one economic argument for why we should not value the lives of older individuals less:
    https://www.thebigquestions.com/2020/04/20/are-old-lives-worth-less/

    For sure, some of the differences of opinion out there can be traced to differences in preferences of trade-offs (and some to differences in opinion regarding facts). That’s true of almost any political disagreement.

    I agree that lock downs of outdoor activities seems like overkill. There’s some speculation (based on past research) that vitamin D is one of the reasons the flu season ends in warmer weather, so being outside is good for the immune system. Also, higher temperatures and outdoor air tends to be more toxic to viruses than indoor air. It seems very unlikely to me that someone would catch the virus outdoors unless they were having a prolonged, close conversation with an infected person without masks.

    I was all for shut downs initially, because of what we saw happen in Italy and some other countries. I thought there was value of information in postponing infections to find better ways to contain, treat, and prevent the spread. And we got a preview of what can happen. New York was bumping up against capacity (fast spread probably due to cold weather and high density), and many nursing homes have had tragic death tolls, but much of the rest of the country didn’t come close to reaching hospital capacity. Although there are some promising early candidates, we also have no idea how long it will take to widely distribute a vaccine. This Fall would be a near miracle. Fall 2021 would still be very impressive by historical standards. Might be even longer.

    Personally, what I am leaning toward now as a preferred policy would be for the federal government to help procure and distribute safety equipment and tests to health care workers and nursing home workers. The focus would then be to separate covid patients from other medical patients (each hospital building would have to declare if they are an all-covid building or all-non-covid building), and keep nursing homes safe. All workers and visitors to those facilities would be tested every day upon entering the building.

    Other than that, I think we can ease outdoor restrictions and let people decide for themselves whether to take risks or not. In some sense, allowing young, healthy people to become infected and develop antibodies is helpful in that it will bring us closer to herd immunity without waiting indefinitely for a vaccine. If Sweden develops herd immunity by the fall and the rest of us are hiding out for another year, we might deem their policy as a success, for instance, especially if our death toll keeps growing to come close to their per capita numbers.

    The caveats I would put on these ideas is that there have been some worrying reports recently about increased stroke risk in asymptomatic/mild symptom covid cases, and some speculation about damage in some infected children. So perhaps we want to learn more about those issues quickly before “opening up.”

    • Dave –

      > I have, however, seen one economic argument for why we should not value the lives of older individuals less:

      I’m struck by how culturally dependent this is. In most of the world, I’d say, older people are valued very highly in society.

      I once had a friend from Africa who, when he first came to this country, said “Is it true what they say, that in America you take old people and put them in homes and keep them from the family? Why would anyone do that? Elders have so much wisdom and value to give.”

    • >> I think we can ease outdoor restrictions and let people decide for themselves whether to take risks or not.

      It’s not as simple as that. People who take risks by not wearing masks are a risk factor to others not completly removed by others wearing masks. It’s not a symmetric situation. The risk of getting the disease is far more consequential than the inconvience of wearing a mask.

      By the way, the same logic applies to large scale issues; death is irreversable while economic hardship is remediable if the political system is willing to provide the means. (So far some elements seem unwilling to remediate on the scale both necessary and possible.)

      • Sure, so maybe you mandate masks in indoor spaces and provide masks to those in need. And perhaps there are large scale events that should still be prohibited to avoid super spreader events. That’s fine. I’m just saying that staying at home indefinitely is perhaps not the best trade off for everyone, and people who are at higher risk can, to a large extent, take fewer risks, especially if we focus on preventing spread in hospitals and nursing homes.

        As far as economic consequences, during the shutdown, some production is delayed and some is lost forever. And there’s no free lunch in terms of just compensating people indefinitely for staying at home and not working. There are consequences, realized today or in the future, for doing so. That doesn’t mean it’s not worth helping people out during these times — it is — but it’s not just play money either. What if there’s no effective vaccine for five years? Or worse yet, what if we have promising, but ultimately ineffective vaccines for the next five years? You could fool yourself into thinking we just need to do this for a few more months indefinitely.

    • Dave said,
      “Personally, what I am leaning toward now as a preferred policy would be for the federal government to help procure and distribute safety equipment and tests to health care workers and nursing home workers. The focus would then be to separate covid patients from other medical patients (each hospital building would have to declare if they are an all-covid building or all-non-covid building), and keep nursing homes safe. All workers and visitors to those facilities would be tested every day upon entering the building.”

      I am leaning somewhat in that direction as well, but would. like more evidence that the proposal would actually do some good rather than faltering because of not considering possible problems with the idea.

      “In some sense, allowing young, healthy people to become infected and develop antibodies is helpful in that it will bring us closer to herd immunity without waiting indefinitely for a vaccine.” This seems very speculative — I think we need solid evidence that this theory would work in reality before doing anything that assumes the theory does work.

      “So perhaps we want to learn more about those issues quickly before “opening up.”” I agree that we *should* to lean more about those issues before “opening up”. But it may not be feasible to learn about them quickly.

      • “I am leaning somewhat in that direction as well, but would. like more evidence that the proposal would actually do some good rather than faltering because of not considering possible problems with the idea.”

        I’m certainly open to considering some problems with that idea. Do you have any in mind? Or are you concerned about “unknown unknowns?”

        “This seems very speculative — I think we need solid evidence that this theory would work in reality before doing anything that assumes the theory does work.”

        The idea that as more people become immune, the slower the infection spreads is uncontroversial, as far as I can tell. It follows from the structure of the epidemiological models I’ve seen. My assumption is that a) people will be pretty good at assessing their own risk and b) their willingness to take risks will follow accordingly. We may not reach “herd immunity” just by allowing people to take more risk because certainly some people will be very careful to not catch it. But R0 should lower as more people are immune.

        “But it may not be feasible to learn about them quickly.”

        If we cannot learn about the possible stroke/children risks quickly, then we’ll just keep that uncertainty in our decision making process. We can even try to estimate when we’ll have that information or how much it would cost to get it, and decide whether it’s worth waiting or not before deciding.

        • Dave said,
          “I’m certainly open to considering some problems with that idea. Do you have any in mind? Or are you concerned about “unknown unknowns?”

          It’s largely the unknown unknowns.

          “The idea that as more people become immune, the slower the infection spreads is uncontroversial, as far as I can tell. It follows from the structure of the epidemiological models I’ve seen. My assumption is that a) people will be pretty good at assessing their own risk and b) their willingness to take risks will follow accordingly. ”

          The questions remain: Are these models good enough to base policy on? Do the assumptions (a) and (b) fit the reality well enough?

          “If we cannot learn about the possible stroke/children risks quickly, then we’ll just keep that uncertainty in our decision making process. We can even try to estimate when we’ll have that information or how much it would cost to get it, and decide whether it’s worth waiting or not before deciding.”

          This might be one reasonable plan to maintain, but it seems it would be wise to consider other possibilities as well.

        • Actually, I realized I misstated my assumption (b). That should’ve been: “b) people can effectively shelter themselves from the risk.”

          But I do think the idea that having more people immune in the community implies slower future spread. It’s hard to picture any mental model of virus spread in which this would not be true, unless we’re questioning whether people actually become immune after infection. Unless you can articulate a specific reason why we shouldn’t believe that premise, I don’t know how else to address it.

        • The stroke thing is probably real, but I’m not sure how much it should change our understanding of the disease – apparently, according to the CDC, the flu also raises your risk of heart attack and stroke! So the question is, is it a very rare thing that is only noticeable above “background” in an incredibly hard hit area like NYC? Or is it likely enough to meaningfully change a young-ish adult’s assessment of personal risk?

          I am very skeptical of the Kawasaki disease in children thing, because some of the New York cases did not actually have COVID. This seems to be more of a case of assuming anything bad that happens in NYC must be COVID-related (which is admittedly a reasonable prior right now, but I can’t really see how someone who never had the disease could get complications from it. False negatives are a possibility, I guess, but still…

        • Dave,

          > The stroke thing is probably real, but I’m not sure how much it should change our understanding of the disease – apparently, according to the CDC, the flu also raises your risk of heart attack and stroke! So the question is, is it a very rare thing that is only noticeable above “background” in an incredibly hard hit area like NYC? Or is it likely enough to meaningfully change a young-ish adult’s assessment of personal risk?

          Exactly. The problem depends on degree. If you are bed-ridden for 1-3 weeks with a respiratory illness, you risk of embolism and stroke will increase. Improper treatment can as well. The data I’ve seen does not make me believe that SARS-CoV-2 poses any major increase for stroke; but said data is nascent and could convince me otherwise with time.

          > I am very skeptical of the Kawasaki disease in children thing, because some of the New York cases did not actually have COVID. This seems to be more of a case of assuming anything bad that happens in NYC must be COVID-related (which is admittedly a reasonable prior right now, but I can’t really see how someone who never had the disease could get complications from it. False negatives are a possibility, I guess, but still…

          This may actually be a problem of false-positives. PCR testing for respiratory virus have ~2% false-positive-rate; antibody-based testing for SARS-CoV-2 seems to have a similar rate. Since these children showed no other symptoms, it is likely they could have been a false-positive for active virus or antibodies.

          This incidence rate makes me wonder if these are incidences of Kawasaki’s Disease *with* COVID-19 (but not from COVID-19). Kawasaki’s Disease has ~20,000 cases per year; so ~60 cases per million of population. That means ~1000 cases per year in NY state. So these cases could be noise and not a real medical event. (We should still investigate them fully, though, to be sure the case is not otherwise.)

          Second, it is not unreasonable to assume that a minute (like 1-in-500,000) children have a genetic predisposition to developing Kawasaki’s-Disease-like symptoms from SARS-CoV-2. (I think it is rare, because we should have seen more cases by now even with the 3-4 week delay in onset.)

        • In extreme hotspots like New York, there are probably incidences of *everything* with COVID but not from COVID.

        • > This may actually be a problem of false-positives. PCR testing for respiratory virus have ~2% false-positive-rate; antibody-based testing for SARS-CoV-2 seems to have a similar rate. Since these children showed no other symptoms, it is likely they could have been a false-positive for active virus or antibodies.

          When ten out of fifteen patients test positive in PCR or antibody testing it seems quite unlikely that it’s just a problem of false positives:
          https://www1.nyc.gov/assets/doh/downloads/pdf/han/alert/2020/covid-19-pediatric-multi-system-inflammatory-syndrome.pdf

        • Carlos,

          Apologies for not being clear. I’m not suggesting all of the positives were false-positives. Rather, I’m suggesting that some may be false-positives and given the low-power of the sample-size, could skew conclusions about prevalence for this mysterious syndrome.

          Per the resort, only 4 patients were positive via PCR: “Polymerase chain reaction (PCR) testing for SARS-CoV-2 has been positive (4), negative (10), and initially indeterminate and then negative (1)” Assuming FPR=2.3% and FNR=20%, the FDR~=9% (range 3-15; see https://www.medrxiv.org/content/10.1101/2020.04.26.20080911v1.full.pdf). So 1-2 of the positives from PCR could have been false positives. Similar conclusion for the antibody-based results. Some, but not all, could be false-positives.

        • 9% of 4 is 0.4 (3%-15% of 4 is 0.1-0.6), Of course 1-2 (or 3-4) *could* be false positives but doesn’t seem very likely according to the numbers you provided. Say there where a couple of false positives (and no false negatives): 8/15 would still be well above the prevalence in the general population suggesting that there is a correlation.

        • Carlos,

          Yikes. I failed at using my calculator on that calculation…batting a thousand today.

          I also mistakenly gave the wrong wrong FDR based on the paper I linked. The positivity of the test 4/15=27%; using NY data provided in the reference I linked, that means FDR has a range of 10-20%; 4*.15 = 0.6.

          So yes, 1 positive may be false (but it also may not be) and the overall results do correlate to some COVID-19 specific effect (whether this is COVID-19 activating Kawasaki’s Disease in the children or something different is still TBD.)

        • Good points, thank you both! I suppose the other hesitancy I would have is: how much long term damage does the disease do in survivors? That’s a hard question to answer on short notice. Some doctors seem concerned about this.

          One other part of my policy preference above that I forgot to mention initially:

          If things don’t go well, you can always re-institute shut downs at any time. For instance, if a new wave accelerates in the fall when the weather gets colder again. It might make sense to make the shut downs more regional rather than state-wide. I think dense cities are at more risk than rural areas, for instance.

  11. Re 1:

    One reason that it’s not as crazy to use deaths rather than lifeyears lost, besides the main one of it being easier data to report on, is that other than the weirdly low under-five mortality rate, the mortality profile of covid-19 across age doesn’t look too different than all-cause mortality across age (https://medium.com/wintoncentre/does-covid-raise-everyones-relative-risk-of-dying-by-a-similar-amount-more-evidence-e7d30abf6821). (The article is for the UK, but I also eyeballed it for a few other countries like Italy and it looks ~ correct).

    So it’s not crazy to think that our intuitive understanding of “a covid-19 death” and “a randomly selected death” is roughly correct. (It does, however, mean we might slightly underrate the badness of mortality causes that primarily or disproportionately kill very young people, like diarrhea, or young-ish people, like car accidents).

    Re 2:
    I feel like this is a cost-benefits thing, and I really don’t see why picnics with members of your household have substantial benefit over eating at home…all the times I can remember doing picnics are for socializing!

    Re 3:
    I think the main thing I’m surprised about is how many people seem excited to promote policies that prolong misery. Like common sense tells you that having an Rt of ~=.95 will converge much more slowly (in the meaningful sense) than Rt ~=.5 (never mind the scarier scenarios where R>>1!), but when we dropped Rt to ~.8 to ~ .9 in most states, people clamored to ease restrictions rather than trying harder to hammer things down further.

    • other than the weirdly low under-five mortality rate, the mortality profile of covid-19 across age doesn’t look too different than all-cause mortality across age (https://medium.com/wintoncentre/does-covid-raise-everyones-relative-risk-of-dying-by-a-similar-amount-more-evidence-e7d30abf6821). (The article is for the UK, but I also eyeballed it for a few other countries like Italy and it looks ~ correct).

      Infants are less likely to have comorbidities. Basically if you look at all cause mortality for people with diabetes, heart disease, obesity, etc who are very overrepresented in covid patients will it also be x times higher for each age group? It will be somewhat higher of course, but how much. I couldn’t find that info when I looked into this phenomenon awhile back. Actuarial tables seems to be available for all sorts of different racial and sex breakdowns but not by pre-existing conditions.

      • Anoneuoid,

        It seems like Mortality Tables for pre-existing conditions — diabetes, hypertension, obesity, etc. — tend to be a private resource of insurers (used to determine premiums for healthy/life insurance and such), who have no impetus to share them publicly.

        • Agreed that we need actuarial tables by pre-existing conditions to be more readily available to the public. This is something that politicians should be looking at right now. Any way to approach this other than by writing to one’s congressional representatives? (In my case, my current House representative would probably lend a deaf ear to such a proposal; but the one out of whose district I was gerrymandered would probably listen — I’m still on his mailing list for his newsletter.)

        • Martha,

          Perhaps you can also email the member(s) of the Task Force directing the response to COVID-19 in your state? They are the ones who are likely making the real decisions; the governors are just enacting them (at least, in theory).

    • Thanks Linch,

      I’ve been in agreement with the Spiegelhalter article you linked to.

      There is another paper that has a similar result from Italy data.

      https://www.lakeconews.com/index.php/news/65261-study-challenges-reports-of-low-fatality-rate-for-covid-19

      Seljak says that getting COVID-19 doubles your chance of dying this year.

      “If you want to know what are the chances of dying from COVID-19 if you get infected, we observed that a very simple answer seems to fit a lot of data: It is the same as the chance of you dying over the next 12 months from normal causes,” said Seljak.

      https://www.medrxiv.org/content/10.1101/2020.04.15.20067074v2

      I would propose another way to think of it is as the average reduction in years of life expectancy by age group. If you’re 95 and get infected, you’re pretty likely to die from COVID-19. Then again, you are about 30% likely to die in the next year anyway. I would say that 95-year-olds have a reduced life expectancy (RLE?) of 1 year. 85yo – maybe RLE of 10 months; 75yo – RLE of 8 months; 65yo – RLE of 4 months; . . . 35yo – RLE of 1 month. This is based on my rough estimations using a US life table by age and averaging for men and women and just increasing the risk of death in a single year and seeing how it propagates down to lower the median survival from that point.

  12. Phil,

    Nice post; thank you for sharing!

    Regarding Point 1:

    Measuring by “life-years lost” relative to age, underlying conditions, and other factors is key to determining how “deadly” disease may be.

    For example, Disease X kills healthy people age 20 at the same rate as people age 80+; meanwhile, Disease Y kills only people age 80+. Over the same period of time, assuming all else equal, disease Y will accrue more total “life-years lost” than disease Y. You can draw the same analogy for underlying conditions like obesity, diabetes, etc., although this is harder because it treads in muddy water relative to HIPAA (so data will be harder to find).

    This allows one to determine how “deadly” a disease is to a given population based on “life-years” lost, which IMO is a more robust measure for this purpose than just “deaths” overall. Of course, “deaths” is still a useful measure and not something we should ignore, especially when ascertaining individual risk of severe outcomes from infection.

    So in sum, both a useful measures but for different ends.

    Regarding Point 2:

    Agreed. In my locale, the governor banned many activities like picnics, but allowed people to shop in packed supermarkets without enforcing distancing or masks for multiple weeks after initiating the shutdown. Of course, my prior being what I, family, and friends have seen; other locales might have been very different.

    This has seemed more like virtue signaling (in many ways, but not totally) than an attempt to actually combat transmission. Sites having high probability of super-spreading are allowed to operate laissez faire for weeks, but someone cannot have a picnic in the park with family?

    That doesn’t make sense in terms virology; the ability of a virus to transmit 6ft in open air under direct UV-light is going to be orders-of-magnitude less than inside a store with stagnant air and no UV-light whatsoever.

    • Twain –

      > This allows one to determine how “deadly” a disease is to a given population based on “life-years” lost, which IMO is a more robust measure for this purpose than just “deaths” overall. Of course, “deaths” is still a useful measure and not something we should ignore, especially when ascertaining individual risk of severe outcomes from infection.

      That seems rather tautological to me. You want to measure deadliness by life years because you think that life years measures deadliness.

      My question is why is it more robust in your opinion? What kinds of decisions would you make based on that information (I”m not talking about on an individual basis for something like end of life care)? What do you want the information for?

    • Joshua,

      Using LYL normalizes for physiological influence on mortality.

      In terms of virology, a virus capable of killing a healthy person age 20 must be far more virulent (e.g., “deadly”) than one only capable of killing morbid or elderly persons. The more LYL (life-years lost) a pathogen can accrue over a given period, the more “deadly” it is.

      Using just “deaths” can be misleading if you do not account for age, morbidity, etc. Using LYL and QALYL _forces_ one to do this stratifying — hence why it is the standard practice for actuaries (so far as I know).

      I’m making no claims beyond using LYL to ascertain the deadliness of SARS-CoV-2 for the population at large and compare it to other known pathogens.

      • Twain –

        > n terms of virology, a virus capable of killing a healthy person age 20 must be far more virulent (e.g., “deadly”) than one only capable of killing morbid or elderly persons. The more LYL (life-years lost) a pathogen can accrue over a given period, the more “deadly” it is.

        Let’s reverse this. Let’s say that there’s a disease that is more likely to kill younger people than older people. Does that mean it is “less deadly” or virulent than a disease that’s more likely to kill older people than younger people? I don’t really get how “deadliness” is some kind of baseline metric.

        Again, I think that the “life-years lost” is a subjective evaluation. I don’t have a problem with it per se – but I do think that there can be potentially troublesome implications. For example, in some communities, elders are highly valued members. There are socially embedded factors here that I think need to be considered.

        > Using just “deaths” can be misleading if you do not account for age, morbidity, etc.

        Yes, I get people saying that – but I’m asking what the underlying logic is as to why the one is more “misleading” than the other – especially since there are so many relevant variables that are hard to consider, and so much subjectivity in how the different metrics are valued. I spoke to some of my questions above:

        https://statmodeling.stat.columbia.edu/2020/05/10/coronavirus-grab-bag-deaths-vs-qalys-safety-vs-safety-theater-all-in-this-together-and-more/#comment-1334308

      • Joshua,

        > Let’s reverse this. Let’s say that there’s a disease that is more likely to kill younger people than older people. Does that mean it is “less deadly” or virulent than a disease that’s more likely to kill older people than younger people?

        No. A disease that “is more likely to kill younger people than older people” is, by definition, a more deadly disease: it costs more LYL to the population than if it only killed elderly or morbid people. The assumption being that young people are much more hardy overall and therefore more difficult to kill — so a disease must be much more virulent on average to kill a young person than an elderly person.

        I’m arguing using LYL to evaluate the potential virulence — i.e., the severity or harmfulness (in terms of physiology) of a disease to a person/population — of SARS-COV-2.

        The points you raise about using LYL to evaluate economic and social effects (while fair) are beyond the scope of what I am claiming.

        • Twain –

          I’ll try this again.

          > A disease that “is more likely to kill younger people than older people” is, by definition, a more deadly disease: it costs more LYL to the population than if it only killed elderly or morbid people.

          It still seems to me that you’re saying that the deadliness of a disease is measured by LYL because deadliness is measured by LYL.

          > The assumption being that young people are much more hardy overall and therefore more difficult to kill

          But that would depend on the disease. Younger people are easier to kill for some diseases. Again, there seem to be a lot of values embedded in your argument.

          > I’m arguing using LYL to evaluate the potential virulence

          But it seems to me that would necessarily depend on the particular disease and wouldn’t be an overall valid construct. And again, you are only picking one metric – age, when there are many relevant metrics that you’re ignoring that could impact on years lost. And then you’re measuring only one metric, years lost, to measure “cost” when there are many ways to measure cost.

        • Twain,

          > It still seems to me that you’re saying that the deadliness of a disease is measured by LYL because deadliness is measured by LYL.

          LYL measures how prematurely a person died relative to their expected age of death from natural causes. It thus allows someone (like actuaries) to measure how “deadly” or “virulent” a disease is based on its ability to kill someone earlier than the should have died naturally. The earlier a disease can kill someone on average, the more deadly/virulent it must be.

          I’m not sure what is tautological about the above.

          > Younger people are easier to kill for some diseases

          Can you provide examples? Almost all viruses, bacteria, fungi, etc., cause much more severe outcomes in elderly people than young people; this assumption is a tenet of epidemiology and medicine.

          > But it seems to me that would necessarily depend on the particular disease and wouldn’t be an overall valid construct. And again, you are only picking one metric – age, when there are many relevant metrics that you’re ignoring that could impact on years lost. And then you’re measuring only one metric, years lost, to measure “cost” when there are many ways to measure cost.

          LYL as a metric is easy to adjust for factors beyond age, like comorbidities, using Standardized Mortality Ratio (SMR); see Andrea Manca’s post above.

          Using age is the easiest because the Mortality Table’s for age are easy to find and access; the values exist for comorbidities, but they are more scattered in literature or are a private resource of insurers.

        • Twain –

          Thanks. I guess I’ll need to think about this some more, but gotta go.

          Just quickly…

          > Can you provide examples? Almost all viruses, bacteria, fungi, etc., cause much more severe outcomes in elderly people than young people; this assumption is a tenet of epidemiology and medicine.

          I was thinking of maybe diarrheal diseases? Meningitis? Malaria?

        • Joshua,

          > was thinking of maybe diarrheal diseases? Meningitis? Malaria?

          I’d have to check about Malaria.

          Bacterial meningitis (this is what most call “meningitis”) is more prevalent in children because [1] daycares, schools, etc., allow for easy spread (lots of unsanitized surfaces and toys) and [2] children usually lack the understanding of hygiene (not chewing on your fingers) that mitigates exposure. Despite these factors, though, bacterial meningitis is very rare — about 1-in-100,000 diagnoses each year — so its LYL total each year is low. Also in terms of pathophysiology after contracting meningitis, I have not seen anything that suggests meningitis kills children more effectively than adults.

          Diarrhoreal Disease is often more deadly in children because of other comorbities that amplify its pathology. The two primary ones being [1] infected children are usually malnourished/deprived and [2] infected children lack access to basic medical care for treatment (like saline IV or potable water to re-hydrate). (You would account for these effects by designating [1] and [2] with an SMR when calculating LYL). Absent of these factors, the LYL from Diarrhoeal Disease would be very, very low — like it is in the US, UK, and Europe. So really, Diarrhoeal Disease is not *biologically* more deadly to children than to adults or the elderly.

          P.S. Writing about things like Diarrhoreal Disease in any context is disheartening. Those deaths are tragic and preventable.

        • Twain –

          I will defer to your better understanding of the issues here. That said, I still fail to get why you want to measure LYL other than that you can do so, and it’s fun to measure stuff, and because it serves some kind of an actuarial function. If someone could explain to me what the actual function of the calculation might be for the LYL calculation with COVID-19, it might help me to understand.

          Perhaps part of what I’m not getting from your specific argument is that it seems to lead to a conclusion that somehow COVID-19 isn’t a particularly virulent and costly disease because it kills a lot of old people and hardly any young ones.

          But having interacted with you a bit now, I know that’s NOT the argument that you’re making. I’m just not clear what the argument your making is.

        • Joshua,

          > Perhaps part of what I’m not getting from your specific argument is that it seems to lead to a conclusion that somehow COVID-19 isn’t a particularly virulent and costly disease because it kills a lot of old people and hardly any young ones.

          This is my conclusion: Because COVID-19 only kills the elderly and morbid, it does not accrue high cost in LYL.

          But — and I think this is where our confusion lies — I am defining “life years” strictly within biological context; e.g., one year of life has the same value for all humans in terms of biology. This is to determine the threat COVID-19 poses to humans as a species/animal.

          It is not for determining the threat COVID-19 poses in terms of costs to society, because ignores the societal, economic, etc., value of the life-years lost of those who died. So I am NOT suggesting its use for ascertaining anything related to that.

          IMO, if we used LYL from the beginning, we would have realized quickly that COVID-19 poses a drastic threat to only a specific subset of the population — e.g., those who live LTCFs, are very morbid, and/or are low SES — and directed all resources to protecting them instead of trying to protect everyone. The analysis forces you to consider age, comorbidity, SES, etc., when calculating LYL, whereas just counting deaths does not per se.

        • Another possibility is that we just haven’t *yet* seen the effects on the middle aged or mildly comorbid etc because they were effectively protected by lockdowns whereas LTCF are inherently hard to protect, AND high risk.

          There are many more people outside LTCF than there are in. The LYL outside if widespread transmission occurs might well be much higher than what we’ve seen so far.

        • Twain –

          > e.g., those who live LTCFs, are very morbid, and/or are low SES

          I have a question about that. We know that it has hit residents at LTCFs hard thus far – but that could be that among older people, those living in those communties tend to have a lot of contacts and live in an environment where other attributes will lead to them getting infected rather quickly. I’m not sure that we know how much the LTCF aspect will scale out once more and more people start “opening up” and getting out into the community and mixing more with infected people… IOW, the LTCF aspect may add speed of transmission that will equalize over time, as well as to some other factors that lead to greater spread and perhaps virulence (e.g., viral load).

          This seems to get back to what Martha was talking about? – disaggregating the age factor and the “congregate living” factor.

          > It is not for determining the threat COVID-19 poses in terms of costs to society, because ignores the societal, economic, etc., value of the life-years lost of those who died. So I am NOT suggesting its use for ascertaining anything related to that.

          As I said to Daniel above, I guess I’m reluctant to go down that path out of concern that it would be used to determine cost to society. I see a LOT of that going on already. And although I can understand the logic to some degree, I also think that it’s highly subjective and envelopes a lot of implicit values (such as that people who live in poorer communities, are of inherently less value) – and I find it a somewhat repugnant. At least to the extent that people aren’t willing to address what the real implications of that aspect are.

          The David Young who has been commenting here told me elsewhere that the disparate impact on poorer communities is “irrelevant.” And unfortunately, I don’t think that’s a particularly unusual viewpoint.

        • Daniel –

          Honestly, I was writing my comment as you were posting yours and I didn’t just copy your thoughts. :-)

        • I will point out here that until all the “survivors” die, computing LYL based purely on the people who died is actually an underestimate. We don’t really know what the long term effects might be – it could well be the case that at the end of the day, many of the survivors lose more years of their life (especially quality adjusted) than the dead.

        • Daniel,

          This could be possible, but depends on specific assumptions being true:
          1. The middle-aged and mildy comorbid were not exposed before lockdown;
          2. They were not exposed during shutdown while visiting supermarkets, big-box stores, etc., while not wearing masks or distancing properly (distancing and masks were not mandated at these locations in NJ, DE, MD for ~2-3 weeks into the shutdowns);
          3. That middle-age alone (I assume age 40-60?) increases risk of severe outcomes (I’m not convinced on this from current data);

          Something that could confound this is if severity of outcomes scales exponentially with comorbidity. That would mean someone whose overweight and grade-1 hypertensive may have minimal risk, but someone who is morbidly obese and grade-3 hypertensive has massive risk. Haven’t seen enough data stratified by specific comorbidities or their grades/severity to ascertain this.

          Joshua,

          Martha made an excellent point about considering “high exposure environments” instead of just LTCFs and congregate living. I 100% agree.

          I don’t like people discussing using LYL to determine cost to society or the economy, either. It is not a tool for determining economic/social costs, but rather the medical risk of how likely something is to kill you before you would die naturally.

          And again, I’m not suggesting using LYL alone to gauge COVID-19. IFR and other measures are useful in their own ways.

        • Twain, we know from seroprevalence surveys that on the order of a few percent of people have had this virus so far. In NY perhaps in the 10-20% but remember that there’s questions there about recruitment/representativeness of the general population.

          On the other hand, we also know that once something hits a LTCF huge swaths of the facility get the virus within a week or two and they stack the bodies outside in the shed. :-(

          what this tells me is that LTCFs had far more prevalence compared to the general society. As time goes along, I expect infections in the general population in the absence of a vaccine to asymptote towards 10x as many as we’ve had so far, whereas among LTCF I expect them to asymptote towards maybe 2x as many as we’ve had so far.

          We can’t say that this “mostly affects LTCFs” we can only say that “it has mostly affected LTCFs so far given that we had a lockdown”

        • Daniel,

          Agreed. I was just posing some points to consider as Devil’s Advocate

          At this point, the lacking leadership by state and federal officials toward more comprehensive and quality studies on seroprevalence is so frustrating — because that data could provide key insight about potential risk as states move to open. The same goes for really ascertaining transmissibility, which still a huge unknown and open debate.

        • > At this point, the lacking leadership by state and federal officials toward more comprehensive and quality studies on seroprevalence is so frustrating — because that data could provide key insight about potential risk as states move to open. The same goes for really ascertaining transmissibility, which still a huge unknown and open debate.

          +1

          Our lack of quality data w/r/t who is getting infected under what circumstances, who is mildly infected vs. seriously infected, etc., is very hard to understand .

        • Twain, we know from seroprevalence surveys that on the order of a few percent of people have had this virus so far. In NY perhaps in the 10-20% but remember that there’s questions there about recruitment/representativeness of the general population.

          I don’t think you know this.

          It is unclear what the prevalence of antibody is in individuals with subclinical or asymptomatic infections and how this assay performs in an asymptomatic population.

          https://jcm.asm.org/content/early/2020/05/07/JCM.00941-20

          Could be the antibody tests are only detecting the most severe 5-10% of the cases, and the rest who had mild/asymptomatic illness have T-cell mediated immunity. There is no direct evidence for that but for the first SARS the T-cell mediated immunity is reported to be much stronger than antibody mediated:

          Animal studies have indicated the importance of T cells in the clearance of SARS-CoV during primary infection and protection from disease [8], [9], [10]. In humans, decreased T cell numbers (lymphopenia) correlated with severe disease, indicating the critical role of T cell-mediated immune response in disease development [11], [12]. While SARS-specific antibody level in SARS-recovered individuals is undetectable at 6 years post-infection, SARS-specific memory T cells persisted up to 6 years following recovery [13]. The long-term persistence of memory T cell immunity could be important in protection against SARS-CoV re-infection.

          Also, for symptomatic SARS:

          Among the cohort, 163 (92.61%) of 176 (χ2 = 200.11, p = 0.000002) were IgG positive, which indicated that most patients who met the WHO case definition were indeed infected with SARS-CoV. As shown in the Table, at ≈7 days after the onset of symptoms, the percentage who were IgG positive was ≈11.80%. This percentage continued to increase, reached 100% at 90 days, and remained largely unchanged up to 200 days. Furthermore, after 1 and 2 years 93.88% and 89.58% of patients, respectively, were IgG positive, which suggests that the immune responses were maintained in >90% of patients for 2 years. However, 3 years later, ≈50% of the convalescent population had no SARS-CoV–specific IgG.

          https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2851497/

          So lots of recent cases may not test positive for a couple weeks. But it looks like we would expect cases from eg, January to be testing positive for antibodies (assuming the response is similar to the first SARS).

        • Certainly, if more non-LTCF people get infected, we will see more non-LTCF deaths.

          But I don’t think the picture of *relative* risk between LTCF and non-LTCF residents will change much (or the individual-level risk).

          The reason I think that is because we have several examples of “congregate” settings with healthier populations (Diamond Princess – mostly elderly but reasonably healthy, Theodore Roosevelt – mostly young adults, etc.) and they don’t show an IFR anything like LTCFs do (2% for Diamond Princess; so far a bit under 0.1% for Theodore Roosevelt, but it’s more recent, though could rise — although I don’t think anyone is in ICU).

        • @Daniel Lakeland: Sorry, I think I misread your comment… if you are referring to “what proportion of all COVID deaths are in LTCFs”, rather than “how much more risk do LTCF residents face compared to members of the general population”, yes, that might change significantly.

          On the other hand, serology does lag (though so do deaths), and I don’t know how many serology studies have been done outside of big coastal cities. I agree that probably less than 10% of the US population has been infected right now, but it might be higher than we’d expect. There might be a lot of un-noticed spread among young-ish people infecting other young-ish people, and nobody getting sick enough to go to the doctor/get tested* until a couple of weeks later it ends up in an older population cluster.

          *One thing I haven’t seen talked about much is the tendency of many American men (in my anecdotal experience, it’s pretty much exclusively a male thing) to avoid doctors even when they really ought to go. Not even because of healthcare costs, it’s more a tough it out / don’t show weakness sort of thing. This might notably reduce the number of people who seek testing.

        • Confused said,
          “The reason I think that is because we have several examples of “congregate” settings with healthier populations (Diamond Princess – mostly elderly but reasonably healthy, Theodore Roosevelt – mostly young adults, etc.) and they don’t show an IFR anything like LTCFs do (2% for Diamond Princess; so far a bit under 0.1% for Theodore Roosevelt, but it’s more recent, though could rise — although I don’t think anyone is in ICU).”

          Good point — thanks for mentioning these examples.

        • Twain said,
          “LYL measures how prematurely a person died relative to their expected age of death from natural causes. It thus allows someone (like actuaries) to measure how “deadly” or “virulent” a disease is based on its ability to kill someone earlier than the should have died naturally. The earlier a disease can kill someone on average, the more deadly/virulent it must be.”

          I some sense, yes, LYL “allows someone (like actuaries) to measure how “deadly” or “virulent” a disease is based on its ability to kill someone earlier than the should have died naturally.” But the real question is: Is this a good measure? Is it the one we should use? If so, why? How does it compare with other measures? For what purposes would it be better (or worse) than other possible measures? Perhaps you personally prefer this. measure to the possible measures — but just because you personally prefer it doesn’t mean that everyone else should accept it as “the” measure. You are free to use whatever measure you prefer — but you need to keep in mind that not everyone agrees that your measure is the best.

        • Martha,

          Apologies — I was not being clear earlier.

          No, LYL should NOT be the only measure/method we use. IFR, SEIR Models, virology, pathology, etc., all provide useful data in their own way that provide the most clear picture of a disease. They all exist for a reason, after all!

          LYL does provide useful context about how virulent a pathogen is to a human (as a biological entity) — a disease capable of killing only those who are morbid or elderly (low LYL), in most cases, is much less virulent than one capable of killing young and old, healthy and morbid, etc., easily (high LYL).

          One way I find LYL particularly (at lest for my understanding) is that, as a method, it forces you to make underlying risks like age and comorbidities explicit during analysis.

  13. The notion than golf flagsticks sitting outside in the open air are a potential vector for disease seems to me borderline outlandish. And yet everywhere that golf has reopened (AFAIK) requires players not to touch the flagsticks… or pick up rakes to smooth sand traps. Is golf a frivolous activity which should be banned altogether? I don’t think so, but it certainly seems no more dangerous than picnics.

    The optics of safety theater carry two risks. The first is that measures which have value which are essentially pure optics might be taken as actually important, causing people to overestimate the risks they are undertaking. If raking a sandtrap is a dangerous act, the so is virtually everything else! While the actual transmissibility of COVID-19 is far from fully understood, the notion than virtually any contact with the outside world should be strictly regulated is obviously wrong.
    Going the other way, though, the second problem of security theater is that measures that are correctly *perceived* as overkill call into question the judgment of the policymakers imposing them, making people lump the really important measures into the unimportant as one more example of things being done out of an excess of unjustified caution. (I admit: unjustified on some metric that is difficult to fully articulate.)
    My personal opinion is that policymakers should not take any actions just to make people feel safer that doesn’t actually make them safer. I’m not saying that’s not easy to do in a situation where what is safe and what is not is highly uncertain. But this is a place where the precautionary principle, whatever its general merits (and I see very few) ought to be sharply circumscribed. People who don’t know what’s safe ought not be making stringent rules about keeping people safe. They obviously feel a decision-theoretic asymmetry here,and I understand it. But it should be resisted. My two cents, with which I’m sure Daniel Lakeland and some others will disagree.

    • Jonathan (another one),

      > Going the other way, though, the second problem of security theater is that measures that are correctly *perceived* as overkill call into question the judgment of the policymakers imposing them, making people lump the really important measures into the unimportant as one more example of things being done out of an excess of unjustified caution. (I admit: unjustified on some metric that is difficult to fully articulate.)

      Agreed!

      The ability of a virus to survive for any extended duration (e.g., minutes or longer) on a surface outside, like a rake of flagpole, in open air and direct UV-light is going to be very small. Most people can ascertain this and may wonder: “Huh, why would policymakers implement such a ridiculous measure?” If this happens enough, they may wonder: “Huh, do policymakers really know what they are doing, or is much of this just for show to make me feel better?”

      If enough people think above, they will lose faith in policymakers both now and perhaps for the future, which could make them less keen to listen now and during future pandemics.

      • 100%.

        Here in Australia, also in lockdown.. I visit the supermarket for food, the one place everyone has to go.
        – No mandatory hand sanitizer on the way in.
        – No mandatory masks, not even on staff,
        – No mandatory hand sanitizer on the way out.

        Some supermarkets have glass screens now between the checkout person and customer. Many still don’t.
        They do have trolley wipes and now even a trolley person now who sprays down the trolley and cleans it before handing it off to you. In one supermarket I saw someone going around cleaning the handles on the fridge/freezer doors. In my regular supermarket I’ve never seen it (maybe they do it).

        I visit the bottle shop (alcohol), also a necessity (trust me).
        Nothing. Belatedly, last time I visited, the guy on the checkout gave a cursory wipe down of the efpos machine after I had used it. No trolley wipes. No hand sanitizer. No masks. No screen to protect checkout guy from customer and the reverse.

        Massive lockdown, can’t go on picnics, sunbake on the beach, meet more than 1 friend.. but the one place indoors with poor ventilation that everyone goes has zero protection.

        And so I 100% agree. Of course, some people are morons, some are dumb and so simple (nonsensical) draconian rules might make sense. But most people are capable of a little thought. The mass of contradictory rules makes people question everything. And even more important many life situations can’t have explicit rules written for them. So people need to make a judgement call. It should be an informed judgement call.

        As a last point, I’ve discussed with a few friends the lack of an edict to wear masks in the shops. “Oh, but there aren’t enough masks to go around, so that’s probably why”. Great point, but instead of pretending they don’t help reduce the spread of covid19, tell people exactly that. Of course WHO didn’t help with their ridiculous ideas and government beaurocracies move slowly, I get that. But here we are something more than 6 weeks into a lockdown that probably costs in Australia $100bn+ a month, and in the US in the order of $1tr a month. This stuff should be screamingly URGENT.

        • Stevec said,
          “As a last point, I’ve discussed with a few friends the lack of an edict to wear masks in the shops. “Oh, but there aren’t enough masks to go around, so that’s probably why”. Great point, but instead of pretending they don’t help reduce the spread of covid19, tell people exactly that.”

          ??What is the evidence that masks don’t help reduce the spread of covid19?

        • Martha (Smith) I think you misinterpreted Stevec. Stevec said don’t PRETEND that they don’t help reduce the spread of covid19, people should be told that they do prevent the spread but that there aren’t enough masks to go around. He’s not saying masks don’t help, he’s saying the opposite.

          That said, the ‘not enough masks’ thing doesn’t really wash. Even cloth masks help some, and there’s no shortage of those.

      • Stevec,

        I’ve seen the same problems as you here in the Northeast US. So have family and friends.

        If the point of shutdowns was the mitigate spread — then the lacking enforcement of masks, distancing, hygiene, sanitizing, etc., at potential super-spreading sites really sends a contradictory message to citizens. (At least where I live; I cannot speak for other states and regions.)

        And I understand some of the widespread anger toward the shutdowns being “unfair”. People see “essential” business — many being large corporations — able to operate without the proper precautions while their local salon/restaurant/store is unable to operate (even though they arguably have MORE incentive to protect their customers because of liability).

    • > People who don’t know what’s safe ought not be making stringent rules about keeping people safe. They obviously feel a decision-theoretic asymmetry here,and I understand it. But it should be resisted. My two cents, with which I’m sure Daniel Lakeland and some others will disagree.

      Actually not at all! I totally agree and am very very much against security theater.

      The big problem here is that there are very clear public health steps that essentially all professional public health people agree with which we should be taking to actually minimize the effect of this virus on people’s health and reduce the duration of economic hardship by making the world relatively safe to go about in.

      WE ARE DOING NONE OF THEM. NONE. NOTHING. NOT AT ALL.

      • Daniel Lakeland said,
        “there are very clear public health steps that essentially all professional public health people agree with which we should be taking to actually minimize the effect of this virus on people’s health and reduce the duration of economic hardship by making the world relatively safe to go about in.

        WE ARE DOING NONE OF THEM.”

        What are these steps?(I can’t read your mind!)

        • Widespread prevalence surveys, ramping up testing capacity, hiring contact tracers, creating means for people to quarantine away from their families to stop spread within households, isolating highly vulnerable populations such as LTCFs as well as the employees of those places, etc. To the extent we are doing any of them it’s at a snails pace. Most we aren’t even trying

  14. We had a bit of this discussion in New Zealand, but definitely a very minor part of the debate. No politician took this up, and much of the response was “you want old people to die!!!” Or a variant of “death panels”.

    Let’s just say the chance that a 60 year old votes versus a 20 year should not be discounted in this kind of analysis.

    • Here’s a very good reason to track life years lost: suppose although the deaths per day are declining, NOW is just the beginning of the peak? Suppose as meat packers and vegetable harvesters and bus drivers and whatever are going back to work, life years lost per day is climbing and climbing even as deaths decline because the people dying are much younger. One 40 year old loses ~ 40 years when they die compared to one 90 year old who loses 1 year. So even if 40x fewer 40 year olds die, the life years lost could be the same. Shouldn’t we track that?

      I think most people are thinking about this wrong, backwards “yeah most of the people who died were old, so this will “deflate” the importance”…. I personally am thinking the other way around… we’re just getting started, we saw the most vulnerable die off, but the real life-year toll is probably still coming.

      • Daniel said, “I personally am thinking the other way around… we’re just getting started, we saw the most vulnerable die off, but the real life-year toll is probably still coming.”

        You may be right. So hard to tell.

        • Considering that people age say 30-60 have been sheltering at home not out and about working… many of them have not been exposed, whereas LTCF residents are exposed whether they like it or not because of the caregivers. But, as we open up and people head back to work… the number of people in middle age who will be exposed will go up, and when a 50 year old dies they lose 20-40 life years, whereas an 85 year old in a LTCF loses ~ 5-7. So even if deaths per day decline by a factor of 2 or 3 if the demographics shift from LTCFs which have already been hit to janitors and bus drivers and meat packers… the life years lost per day could easily increase even as the lives lost decrease.

        • Start looking into the percent of deaths due to:

          1) Inappropriate medical treatment
          2) Avoiding medical treatment due to #1

  15. Phil,

    From a societal standpoint, your life is worth *more* than that of a 20 yr old.

    You have accumulated knowledge and experience that are valuable to society. Of course all of your years are important to you, but they are also important to society.

    Twenty-year-old is still a blank slate. Whatever the apparent potential in any individual, none of it is yet realized. The twenty year will need between one and two decades of seasoning to become an asset to society. It’s nice that child activists to babble about their favorite cause, by why anyone would listen to a child is beyond me. Children don’t know anything. Even a PhD student – loud as h/her mouth might be – is probably short on knowledge and definitely short on experience. It takes a long time in life for us to learn how stupid we are and what that means for what we actually know.

    The loss of a typical 20 yr old leaves a hole in the future. But the loss of a typical 35-55yr old leaves a big hole NOW.

    that’s my two cents.

    • Jim said,

      “Phil,

      From a societal standpoint, your life is worth *more* than that of a 20 yr old.

      You have accumulated knowledge and experience that are valuable to society. Of course all of your years are important to you, but they are also important to society.”

      I’m twenty or so years older than Phil. Does that mean that from a societal standpoint, my life is worth more than his? ;~)

      • Martha:

        From our perspective here, your worth is determined by the number and quality of your blog comments. Your Blog Quality Adjusted Life Years (BQUALY) number is high, as is Phil’s.

        • I know how to pronounce QUALY. How does one pronounce BQUALY? To rhyme with “equally”? Or with a quick-and-difficult-on-the-tongue transition from “b” to “k”. (Any linguists around to give the proper lingo on this?)

        • Thanks — although I think it may take a bit of practice before I get the knack of it. From a YouTube video on clicks ( https://www.youtube.com/watch?v=Nz44WiTVJww) I find that there are four different clicks in at least some languages. (I think some of them may be related to what I once was told were labial, lingual, and uvular trills. I once learned to do all three of those at once.)

      • As long as you’re teaching people you’re part of the critical experienced middle, right? :)

        When you get to the point when you’re spending all your time slogging martinis in the golf cart, then you’re expendable. :)

    • jim, your calculation depends on all kinds of assumptions, including especially discount rates to be applied to productive economic activities. It seems with such low rates as we’ve seen the last decades, there’s no way the present value of the pure economic production of a 20 year old is less than Phils.

      • As I pointed out, the avg 20yr old has more long term value. But the 50yr old has much higher – and perhaps critical – short-term value.

        Oh, yeah, now I remember where I heard this. Recently I listened to a series of lectures on the history of Africa. According to the lectures, one devastating aspect of the HIV epidemic in Africa is that HIV tends to kill people just as they reach the prime of their careers and thus has a staggering effect on economic growth in Africa.

      • Yeah, thinking about it more I see a big problem in your position:

        People’s long term economic productivity depends strongly on having experienced people to learn from.

        If you cull the 50yr olds, then the long term productivity of the 20yr old declines. The longer you keep removing people from the middle, the slower the younger people will learn and the longer economic productivity will decline.

        How many consulting firms do you know of with only 25yr olds? :) None, right? Because it doesn’t work. The 25yr olds need experienced people to guide them.

        • No doubt that killing off *all* the 50 year olds would devastate a lot of businesses. But then, killing off all the 20 year olds would do the same several years out, when there’s no one to age into the responsible positions.

          at the margin though, when we’re talking about losing say much less than 5% of the population of 20 or 50 year olds, the 20 year olds are way more valuable economically due to the duration of their continued productivity and the small discount rate.

          I don’t doubt that the effect you’re talking about is a thing, just that at the margin it isn’t big enough to make the calculation switch.

        • I’m sure you can say the effect is smaller as the number of fatalities decline, but I don’t think you can say it’s nonzero. Seems to me it simply scales with the number of people. There’s only so many people an individual can mentor.

        • This whole discussion strikes me as well down the rabbit hole. You are debating the relative values of different age group’s lives but seem to have lost sight of the underlying issue of how should “value” be measured, including “value to whom?” Even if it is implicit that you are trying to gauge the value of one year of life, on average, to “society” (i.e., excluding considerations of value to the individual or their families) of a 20 year old vs a 50 year old, myriad questions present themselves. What are you assuming about the race and gender of the people? What are you assuming about their underlying health conditions? What are you assuming about the unintended consequences of the policies enacted to protect some of these lives (e.g., higher suicide rates due to job insecurity, fewer highway accidents due to shelter-in-place rules, etc.). No doubt there are entire research programs that could be devoted to exploring this rabbit hole.

          I don’t mean to dismiss these considerations by casting them as a rabbit hole. What I mean to point out is that there is a danger of losing sight of basic ethical questions. I don’t take the utilitarian framework for granted – it can be valuable when applied at the right time. But it is dangerous if adopted too readily. Would you have a physician decide whether to treat a 75 year old’s heart condition or donate their time to helping an underprivileged 20 year old avoid future high blood pressure? Somehow we allow those choices to be made without intervening.

          The difference is that we are considering policy choices. Our policies affect the physician’s individual decision (through various policies, including licensing, reimbursement, insurance requirements, etc.) and we are now – under dire circumstances – debating how best to limit the spread of COVID without unduly destroying the lives of the uninfected. I don’t pretend to know how we should set policy, but I am leery of applying a utilitarian framework without considering alternatives. That is the rabbit hole I would like to avoid.

      • Look even in the tech industry, where knowledge is exploding, the industry is lead by a cohort of very experienced people (like Andrew for example). Had all those people been killed in an epidemic or a war in 1995, the younger people in the industry would be much less productive.

        • Again, you’re basically confirming the nonlinearity of the effect. If you lose 0.5% of all people age 50 vs 0.5% of all people age 20 it’s true that maybe each day of a 50 year olds life is economically directing resources at 2x the rate that a 20 year old is… but they’re doing it for maybe 15 more years, whereas a 20 year old is going to be doing it for 45 years and 30 years from now they’ll be a 50 year old, they’re not going to stay a 20 year old their whole life.

          So, you have to somehow argue that the present value of the 15 years that the 20 year old will spend being between 50 and 65 is worth essentially zero today due to discounting AND that the 30 years they’re going to spend being between 20 and 50 is so non productive that the 50 year old wins out because their productivity from age 50 to 65 is discounted so much less that it beats out everything the 20 year old will ever do in their lifetime.

          Do you understand what the implications of that assertion are? It literally suggests that there is zero value in having children ever. with those kinds of discount rates it couldn’t possibly be worth it for society to afford the 20 years of child care and education it takes to get someone to be a productive 20 year old. That kind of discounting could only ever exist in a world where say an asteroid was known to be coming and wipe out society exactly 15 years out.

  16. It’s been done, about 10y per death, though this is a slightly simplistic calculation that overestimates, probably substantially.

    As for fairness and spreading the pain, taxation is a good starting point. Good luck with that.

  17. About counting years of life lost.
    I get the reasoning regarding age. I buy the notion of a fair life-span, four score and ten, or whatever. So the 80-years old has gotten her/his fair share, and the 20-years old has not. But that doesn’t hold when it comes to life expectancy differences by race, socio-economic status, or disease status. Nothing fair about that (imo). Then counting years of life lost will only serve to justify/reinforce the status quo, even to blame the victim.

    • Oh no, there are plenty of stops on the way.

      Value to society is already used when we determine who goes on death row. It’s not like it’s one step from there to eugenics.

      But also, I think you may have missed the point of the stuff about the number of days left. If you value the life of a 90-year-old equally with that of a 20-year-old, you’re saying a day in the life of a 90-year-old is much much more valuable than a day in the life of a 20-year-old. (That’s a true statement no matter who is deciding what ‘valuable’ means. You don’t seem to like ‘society’ determining the value, but hey, somebody’s gotta do it. Are we supposed to say “hey, let’s have Thomas decide?”)

      If your argument is that I’ve chosen my words poorly, go ahead and suggest other words. If, instead, you’re suggesting everyone’s life is ‘equally valuable’, or that we should somehow make decisions as if that’s true, I think you’ll find that that position doesn’t really stand up to scrutiny.

      • Phil, your words are fine.

        I agree with you on age as a criterion for prioritizing care, but from an argument of fairness, not because the young are more useful to society.

        I don’t think value to society is the key argument. Ethical frameworks are multidimensional, and the devil is always in the details. If resources are lacking I would first ask why (lets buy another respirator!), if that’s out I’d consider all the ethical arguments (autonomy, justice, beneficience,…), and if that doesn’t help I’d consider a lottery.

        It’s fine to disagree.

        • A lottery has a lot of things to recommend it, but not if everybody gets an equal number of tickets. If firemen pull up in the ladder truck at a building that is being consumed by flames, and they can either drive to one end and save a 90-year-old with dementia and pancreatic cancer, or to the other end and save a healthy 12-year old, but they can’t save both, I think they should save the 12-year-old. Do you really disagree with that?

          Of course you’re allowed to disagree, but if you disagree then I’ll say our moral codes are so different that we are going to have to ‘agree to disagree.’

  18. Phil, your musings seem to switch perspectives in an odd way. By this I mean, you seem to be alternately thinking in terms of the community perpective versus the individual perspective.

    Your first topic, deaths versus life years lost, seem to be couched in a community perspective. From that perspective, what is lost to “us” when someone dies is the number of years they would have been expected to live. So of course it makes sense from that perspective that losing 70 expected life years (from losing a 20 year old) versus losing 10 expected life years (from losing an 80 year old) weights the tragedy towards the former. Yet, I know a number of 80 year olds who are value their own life as much now as they did sixty years ago. So from the individual’s perspective death is no less or perhaps only a moderately less tragic at 20 then at 80. Rather you could make the case from the individual’s perspective that death in infancy or death in advanced senility is less tragic then death at a point when one’s conciousness can fully grasp the weight of losing ones life. So given that individuals perspective, I do think that measuring deaths is way more meaningful than measuring life years lost. There are a great many elderly people who see the age trends in the death rate and rather than thinking “well that’s good at least the children have a good chance of making it out of this okay” are thinking (in a gentle mocking stereotype of my own mother’s capslock fondness) “WHY THE HELL AREN”T YOU PEOPLE WEARING FACE MASKS!? I DON’T WANT THE CORONA!”

    Conversely, your musing on safety versus safety theater seems to way more couched in the individual rather than the community perspective. The individual might rightly ask “why can’t I picnic in the park?” Seeing as how, all else being equal, an outdoor picnic with ones own loved ones is not that different in terms of interactions than a walk in the park. The problem from the community perspective is that by permitting picnic we no longer satisify the assumption of “all else being equal.” If picnicing is permitted that has the possibility of increasing the density of users in the park by both introducing additional activities that will draw additional visitors and also by allowing an activity that has increased residence time in the park for each small group of individuals that partake in that activity. The risk to the community is in this risk of increased density. While it may be “theater” in the sense that yes, other activities are being permitted that probably increase risk more than picnicing and, yes, picnicing probably only carries marginally more risk than walking in the park, I would argue that it is effective theater in transmitting this community signal: picnicing in the park is a luxury which carries some small amount of inherent risk and for the sake of the community we should be willing to sacrific luxury to reduce the risk to the community as the whole.

    Regarding point 3: you’re right, we’re not all in this together and we have no coherent plan for how to get out of this together. Many people are going through economic hardship right now. They’re being told by the government that they can’t work and they’re getting the bare minimum in support to maintain their livelihood. Unfortunately, I don’t see any workable scenario where we can somehow open the economy without risking longterm economic damages that will be onpar as what we get with an extended shutdown. There are a whole lot of people who aren’t willing to sacrifice their 5 or 10 or 20 years of expected life to increase the expected quality of life for someone with 50 or 60 years of expected life. So, the economony will have less activity even if all business were allowed to operate. We’ll also have alot more dead bodies to bury, which does tend to put a damper on things. On the otherhand, I also don’t see any workable scenario where we fully embrace the community perspective of shared sacrifice and looking out for your neighbor at least not until an election or three.

    • Dalton,
      If you think the switching of perspectives exemplified by this post is ‘odd,’ you’d have been gobsmacked if I had added the two or three other topics I had intended to add before realizing the post was getting too long already. I am an individual who is a part of a family and part of society; I have elderly parents; I have friends ranging from about 20 years old to about 85; I have friends who have lost their jobs due to the pandemic response; I have friends who have lost loved ones to the novel coronavirus. I have not just one perspective, or two or three, but dozens at the same time. I contain multitudes.

      The fact that I want to see reporting of life-years lost does not indicate that I’m indifferent to the deaths of old people. I absolutely am not. Both of my parents are still alive and I hope they stay that way for a long time. My dad, in his mid-eighties, is struggling enough day to day that I think even he would agree that some ‘quality adjustment’ would be appropriate when counting his days compared to those of an average younger person, but my mom is a living breathing illustration of the fact that one can have a rich and happy life in old age. Everyone should wear their damn masks. But I still want to be able to see life-years lost, not just ‘deaths’.

      As for picnicking in the park, I guess I’m missing your point about community vs personal perspective. For one thing, I have no particular interest in picnicking in the park, I’m just using it as an example of a restriction that seems to me to be mostly safety theater rather than actually promoting safety, although I admit that I might be wrong about that. But, in the sense that I don’t care about this personally, I’m still looking at it from a community perspective, I think. Anyway I agree with you that one issue with picnics might be that they implicitly signal that things are more normal than they are; I tried to touch on this in my post when I said ‘maybe that gives people a false sense of security and would nudge people towards dropping their guard on more consequential behaviors’.

      Point 3, sure, if businesses were allowed to operate we would still have a major recession and we would kill more people. As I mentioned, look at Sweden. As in so much else, it’s a question of tradeoffs. Where do we want to be? But also: whatever economic hardships we suffer, under whatever rules we follow, some people are going to be hurt much more than others; indeed some people will benefit economically. I think the degree of unfairness is excessive and I’d like to see it mitigated. I just don’t know how that could happen to a significant degree, realistically. Maybe it can’t.

      • Phil, I agree with you about the multitudes of perspectives. I too want to see a *lot* of different “ways of measuring” whats going on, so I can decide which ones are most useful.

        As for picnicking, I think at least in places like the bay area and LA county, if you explicitly say “picnicking in the park is allowed” within hours there will be LARGE congregations of people, because they’re all looking for something to do that’s nicer than being locked in their apartment. Here in Pasadena area they opened Eaton Canyon trails, and it was JAMMED with people, they had to close some of the parking to try to keep the density down.

        If you want to enable some of this stuff, I think you’d do well to have a website where you can put in your address and some demographic information (average age of person in your household or something) and it assigns you a random number that tells you which days of the month you can go to park facilities, and then you’d evenly spread the occupancy around the parks, and also reserve some days of the month only for the elderly etc.

        It’s not so much about it being dangerous to go to park, but it’s dangerous to have a lot of correlation in when everyone wants to go to the park (like everyone wants to go saturday between 10am an 3pm etc)

      • Thanks for the reply Phil. I didn’t mean to imply that your are indifferent to the deaths of old people. Rather I wanted to point out that life years lost fails to capture the inherent loss of a death. I don’t mean this in some idealistic “every life is precious and every death is tragic” way (although that is in some sense true). I mean it more in the way that death is frightening to most individuals of any age. I go so far as to say it is equally frightening. And I think that fear operates in a way that is much immediate. I’m 37. I fear catching coronavirus and dying. But when I think about my death, I’m not particularly afraid of losing that boat I vaguely plan on buying in retirement in 30 years. I’m afraid of not being able to hug my wife tomorrow. For this reason, I don’t think the fear of death declines linearly at with age.

        This is a poor use of a technical term, but life years are not exchangeable. Two deaths with an expected life years lost of 20 years does not equal a single death with an expected life years lost of 40 years. A single death carries with it a certain amount of mass irrespective of age. Yes we lose the economic potential of productive life years for someone younger. But for me that concept is so abstract as to be essentially meaningless. When I think about my father who died 15 years ago around the age of 50, I don’t think about the loss to the economy or to society, I think about how he was not able to attend my wedding or how he won’t be able to know or hold his grandchild. When I think about my grandfather who died a few years ago at 94, I think in much the same terms: he won’t be able to know or hold his grandchild. That connection is what is loss to our family.

        We could also seek to quantify deaths in terms of grief for those left alive. Is it any less traumatic for a parent to lose a child at age 15 versus age 45? Or for a child to lose a parent at 50 versus 75? For the second case, yes, probably, a little, but not linearly. Grief is certainly a process that differs among the bereaved, but I think we can agree that it is more acute immediately following death and decreases over time. We can’t amortize it over the expected life remaining for the person who has slipped the surly bonds.

        For these reasons, I think that deaths is a more appropriate measure even if the elderly are more at risk of dying.

        • Dalton,
          Losing your father that young is a cruel blow and I’m sad for you.

          You are welcome to bring ‘economic potential’ into the picture if it suits you, but it certainly isn’t necessary. Your father dying 30 years ‘early’ cost you many experiences with him, and it’s not just you but also the rest of your family, and all of the friends he had or would have had. Surely the death of your father around age 50 was a bigger loss to the world at large than the loss of your grandfather at 94, no matter how great your grandfather was. And if, the year your father died, there was somehow a choice for the death to happen to you, or your father, or your grandfather, I’m guessing all three of you would have chosen to say goodbye to granddad.

          If you imagine a choice between losing 10,000 randomly-selected 50-year-olds (losing them to death, I mean) vs 10,000 randomly-selected 94-year-olds, are you really indifferent between these, or even close to indifferent?

          If so, well, you’re entitled to your view of course, but we see things very differently.

        • As with many things, there are many ways to look at death, life, value of life, etc. When considering the value of a life, I tend to think of Herodotus’ saying, “Count no man happy while yet he lives”. For example, when my father died (at an age slightly younger than my current age) my salient thought was that he had died quickly, in his own room, in his own house, that he had designed himself, and built with is own hands. For him, I think that was the best way to go, far better than lingering in a hospital connected to tubes and other paraphernalia. But I can’t say that that would be the best way to die for everyone — what constitutes a good death, and a life worth having lived, varies from person to person, depending on what matters most to them.

  19. If your model for engaging with the epidemic is mitigation, aka “flatten the curve”, with an understood “everyone gets it eventually”, then the Swedes needn’t care about their death rate NOW because they predict that everyone else will eventually catch up.

    The counterargument is that if we delay these deaths, we may have less of them because
    — we make fewer mistakes in treatment because we learn about the disease
    — we find better treatments (e.g. effective drugs)
    — we manage to cut off the tail of the curve with a vaccine
    — we can achieve containment

    Picnics are a problem because they increase population density in the park, and that’s loosely associated with increased transmission. A sunny day can turn a picnic in the park into a mass event. If you want, you can have a picnic in a solitary place; that’s a) epidemiologically ok, b) police won’t find you there, and both effects are correlated.
    It’s like, “why can’t I have my solitary walk on the beach?” You can’t if lifting this restriction results in a mass event at the beach.

    and we are in this together. Containment is something we can only achieve on a social scale, if enough people make the individual choice to not care about that, we can’t contain. That’s why China was able to do it: because people in that society are used to making individual choices that benefit society, so the lockdown worked. (Someone described this as a property of Confucian societies.) You get into the paradoxes of the common good here: if I go to work, I benefit individually, but if we all go to work, the damage that society sustains (and that we all must bear) becomes greater as if nobody (nonessential) went.

    I think the big thing is that if we get these kind of pandemics more often in the future, i.e. if we have another one in ten years, and we’re saying “we don’t care about people older than 70”, then the people whomare working towards their retirement will stop doing that earlier in life, because they can no longer be assured that they’ll enjoy it. Think about the economic effect of that! Think about the utility of lowering social cohesion as a society. Aas long as we are a society, we are always “in it together”, because social norms governing individual choices result in social outcomes that have individual benefits, even if breaking these norms might have a momentary benefit for the individual. If you teach a money-utlitarianism that doesn’t consider social/societal effects in its framework, you’ll end up at a dysfunctional society.
    The US has just thwarted a worldwide ceasefire for the duration of the pandemic at the UN level on Friday. To me, that’s evidence of dysfunction.

    • Sweden is probably right in at least that containment will not be successful for most mainland nations*, and that a vaccine will come too late to make much difference. (The often mentioned 12-18 months is about how long historical flu pandemics tend to last anyway. Of course, this isn’t a flu virus, but… In 2009-10, the vaccine became publicly available after the second wave had already peaked.)

      Improvements in treatment might make some difference, however.

      *The places that look to be likely to succeed are either islands or practically-island isolated areas like South Korea.

      The point about retirement is good. I don’t know about lowering social cohesion, though. It generally is true that we consider deaths among the young to be “more tragic” than those among the elderly; that’s not new, so I don’t think that viewpoint inherently weakens social cohesion. Outright de-valuing those lives would, but I don’t think talking in terms of life-years reaches that point.

    • It’s funny how the “lockdown early, lockdown indefinitely” advocates pitched it as a short-term measure needed to “flatten the curve” right up until the curve appeared to flatten. Now they seem to be switching tunes to “it wasn’t just about flattening the curve”.

      Sigh.

      • Brent –

        > Now they seem to be switching tunes to “it wasn’t just about flattening the curve”.

        There’s obviously a political aspect to practically all the discussion of the government mandated shelter in place orders, but I have to say I’m surprised to see such naked simplistic spin from someone who as far as I can tell is a regular contributor at this site.

        Trump’s shelter in place mandates were informed by the same public health officials who established the criteria for the lifting of the mandates. Those criteria are w/r/t trends in things like hospitalizations, deaths, and infections. All along the rationale put forth was to prevent overwhelming hospitals and healthcare workers, and to not exhaust needed supplies – but also to buy time to build out a robust infrastructure for testing/tracing and isolating.

        There is no inconsistency.

        Using your approach, we might as well say that the “openists” pretended to go along with Trump’s shelter in place mandates only to the point where they could gain enough political foothold to advance the aim of establishing herd immunity so that they could get away with letting a bunch of old folks and minorities die so they can get on with their business.

      • Brent –

        > Now they seem to be switching tunes to “it wasn’t just about flattening the curve”.

        There’s obviously a political aspect to practically all the discussion of the government mandated shelter in place orders, but I have to say I’m surprised to see such overtly simplistic spin from someone who as far as I can tell is a regular contributor at this site.

        Trump’s shelter in place mandates were informed by the same public health officials who established the criteria for the lifting of the mandates. Those criteria are w/r/t trends in things like hospitalizations, deaths, and infections. All along the rationale put forth was to prevent overwhelming hospitals and healthcare workers, and to not exhaust needed supplies – but also to buy time to build out a robust infrastructure for testing/tracing and isolating.

        There is no inconsistency or goalpost moving.

        Using your approach, we might as well say that the “openists” pretended to go along with Trump’s shelter in place mandates only to the point where they could gain enough political foothold to advance the aim of establishing herd immunity so that they could get away with letting a bunch of old folks and minorities die so they can get on with their business.

    • “That’s why China was able to do it: because people in that society are used to making individual choices that benefit society, so the lockdown worked. (Someone described this as a property of Confucian societies.)”

      I think maybe in China you do it or else.

    • I somewhat disagree that we would get there anyway is the Swedish response. They didn’t expect it would be so difficult to isolate nursing and care facilities. They’ve said so. I doubt they would have ‘locked down’ the country but I think they would have treated nursing homes differently, with more resources thrown at what has turned out to be the death story. In MA, for example, 65% of all the dead are 80 and over, 86% are 70 and over, and about 60% are from nursing homes. That means 4350 deaths age 70 and above, with about 3000 from nursing homes. (I’m assuming a very high overlap between age 70 and up and nursing homes.) If we had thrown resources at one group, we’d have under 1000 deaths (650 in raw comparison). It’s such an obvious target.

  20. I absolutely agree with the life years lost. I think part of the reason why there doesn’t seem to be a public debate about this is that it’s hard to convey the statistical nature of the concept. People seem to always want to take it on an individual or even personal level (“Are you saying the life of my grandpa is worth less than yours?!”), while it’s only meaningful for a population, e.g. to compare countries or different pandemics. This is besides the fact that the number of deaths is already statistical in nature – media likes to portray it as hard and fast, but in truth we have no idea how many people really die of corona (or with it), it’s just a best guess depending on a (necessarily) very flawed reporting system

    QALY or DALY would be useful if we were seeing lots of disability as a result of Covid, which doesn’t seem to be the case. But we should be aware that reporting LYL is equivalent to setting the quality adjustment to a constant 1, which can be a reasonable approximation, but should be a conscious choice. Just as reporting deaths instead of LYL is equivalent to assuming constant expected life years left, which never seems to be a reasonable approximation.

    • Perhaps this is the article: https://wellcomeopenresearch.org/articles/5-75/v1. (I don’t have a subscription either.)

      I’m not sure why this paper strayed from the standard practice of using standardized-mortality-ratio to represent risk from COVID-19 (see Andrea Manca’s above comment). So comparing their methods to others is difficult.

      Further, they do not account for severity of comorbidity; which can have a strong effect on SMR and subsequent LYL. For example, obesity of BMI2 for BMI>36.

    • Konrad, thanks for pointing out that graphic, I think I’ll do a blog post about it and invite people to come up with a better way to display the same data.

  21. Has anyone calculated the life-years lost from the 2009-10 pandemic? I’ve read that had a fairly “flat” mortality curve by age. So while there were relatively few deaths, the life-years lost might be larger than expected.

  22. Regarding research of the folks who did the Santa Clara study….

    OK. Now this is classic:

    > Just 0.7% of Major League Baseball employees tested positive for antibodies to Covid-19. The small number of positive tests, came as the league continues to plan to start its delayed season.

    […]

    “I was expecting a little bit of a higher number,” Bhattacharya said. “The set of people in the MLB employee population that we tested in some sense have been less affected by the Covid epidemic than their surrounding communities.”

    ——

    So I expected a result based on my theory, and I tested my theory and got a different result than what my theory predicted.

    Obviously there must be something wrong with the sample I used because they didn’t match the results I expected.

    I mean, it’s not like there could be something wrong with my theory.

    https://www.google.com/amp/s/amp.theguardian.com/sport/2020/may/11/mlb-baseball-covid-19-coronavirus-antibodies-test

  23. Phil:

    Regarding (3), taxes would probably work to some degree. We can increase taxes for those that have “won” from this pandemic, like Amazon (as Saez and Zucman have advocated for: https://www.nytimes.com/2020/03/30/opinion/coronavirus-economy-saez-zucman.html). Increasing taxes on those like you and me who have not been negatively impacted (financially) from the virus is also reasonable and then you’ll have more to distribute to those that have been most affected.

    • Michael J said, “Increasing taxes on those like you and me who have not been negatively impacted (financially) from the virus is also reasonable and then you’ll have more to distribute to those that have been most affected.”

      +1 — (at least if the increased taxes indeed are distributed to those who have been most affected by the virus)

  24. Let’s assume we knew the IFR given age (down to a single age) and sex (unconditional on comorbities) — just say that for anyone over 30 it is equal to the annual mortality risk from a 2016 actuarial life table. There is mounting evidence that this is not at all unreasonable assumption. We also have population tables by age and sex.

    IFR_a_s = IFR for age ‘a’ and sex ‘s’

    Base_rLE_a_s = median remaining years of life for age ‘a’ and sex ‘s’ based on life table (all-cause mortality)

    US_pop_a_s = population count in US for age ‘a’ and sex ‘s’

    Then an estimate for total Years of Life Lost (YLL), if everyone over 30 were to get infected/exposed (and assuming negligible deaths under age 30), would be:

    sum over s [sum over a [ IFR_a_s * Base_rLE*_a_s * US_pop_a_s ] ]

    What if this turned out to be around 100 million years? What would we do with that number? Since there are about 200m Americans over 30, then on average this is half a year of life lost per person. Is this relatively good or bad? Relative to what — the 1918 flu? Seasonal flu? Cancer? Smoking?

    It’s an interesting number to know, but still we’d want to have that same calculation for other causes of mortality to compare with.

      • Thanks!

        Thinking a bit more about it, maybe under our political/economic system of values (and not, say, Tibet’s), the YLL should really be changed to YWLL – Years of Work-Life Lost, or something. If we could figure out the total years of work-life lost (assuming some things, like maybe assuming everyone retires at 70), then that could be used in an economic argument. We often hear things life “Every year the flue costs the US economy $10.4B”.

        https://www.cnbc.com/2017/10/30/the-flu-costs-the-us-economy-10-point-4-billion.html

        Although these calculations are probably not considering YWLL.

        • Yeah, I think those calculations “the flu costs X amount per year” are based on lost productivity and time off work, not deaths.

    • statflash,
      Sure, the same is true of deaths, too. Is, say, 100,000 deaths a lot or a little? About 3 million people per year die in the U.S. so (for instance) 100,000 seems like a drop in the bucket, especially since some of those people were going to die this year anyway. On the other hand, it’s about 2.5 times the number of people who die in car accidents every year, and that number of car accident deaths is thought by most people to be tragically high.

      To take your hypothetical question, what if we end up losing 100 million years of life, how big is that? Well, 400,000 Americans died in WWII, most of them pretty young. If they lost 55 years each, on average, then that’s about 25 million years of life lost. Your hypothetical situation would be a loss of life-years about 4 times as high as WWII. I think we could agree that’s a big number, even in our more populous country.

      Another approach that I like is Daniel Lakeland’s suggestion to divide by 80 to get a rough number of ‘lifetime equivalents’. 100 million years, that’s the number of years in about 1.25 million lifetimes. Again, it’s pretty obvious that’s a big number.

  25. Just want to preface that I am a microbiologist, and am really just wondering about some of these issues, and am confused.

    I have been wondering about some of the issues you describe. To me, as a biologist, it seems as if most of the news coming out about the virus is positive and I would assume a basic bayesian approach would be used by those leading the charge which would cause prior updating, but I don’t see it.

    In March perhaps our prior related to mortality was somewhere between 1 and 3.4%, but we should be updating it, and this should be reflected in the media: 1.) Drugs which decrease time to convalescence (although limited supply should tick mortality down a bit because it reduces hospital usage), 2.) Many antibody studies suggest mortality rate well under one, 3.) Vitamin D modulation of severity is good news during summer time (people will get more vitamin D), 4.) There is likely some reduction of transmission in hot-humid air, 5.) More insight into differences with 1918 should create optimism (my friend who is an emergency physician sent me an update for emergent medicine, and as of a week or two ago only 3 children had died in the entire US; the mortality tables suggest someone over 80 might simply have an increased mortality rate of someone a few years old if exposed to coronavirus) All together this seems to suggest some optimism. I don’t think we really see any of this in media coverage. To me this has broadly suggested political bias. If you’re a liberal and or an intellectual, then part of membership of those groups mean you have a very conservative take on the epidemic.

    There was an Aeon piece a week or two ago by a bioethicist claiming we are essentially asking the young to take a hit for the old, and we aren’t talking about that at all.

    I just find it all bizarre.

    There will be an argument like: we may have a second peak in winter which is bad because it will overlap with the flu.

    Okay, so why not increase transmission during the summer, when we don’t have to deal with the flu and people will have less vitamin d deficiency.

    I don’t know if I am just crazy or what.

    • Charlie,
      It’s very widely acknowledged that old people are at far more risk than the young, but the economic pain is borne by everybody, maybe especially the young on average. I don’ agree that “we aren’t talking about that at all.”

      There’s nothing remotely unusual about the young being asked to bear most of the burden. The people going off to fight and die in WWI, WWII, Korea, Vietnam, Afghanistan, and Iraq were mostly young. Of course, the ones who survive get to grow older and someday it’s their turn to be in the politically powerful generation. I wouldn’t say this makes it fair.

      I never heard estimates of IFR as high as 1-3.4% for the general population. If you were hearing numbers that high in March then you were seeing different sources from those I was seeing, which is not surprising since you’re a microbiologist. I was seeing numbers roughly 1/10 that high, something like 0.1-0.8% for members of the general population who receive good medical care. I don’t think I ever saw an estimate as high as 1% except for the elderly; for the very old it seems to be much higher than that, more like 10%.

      I suppose whether you see the glass as half full or half empty depends on how full you used to think the glass was. I’ve always assumed treatment would improve — indeed I assumed one of the intended side-effects of the shut-down was to give us time to develop improved treatments before the number of cases got really high — so your #1 was already built into my assumptions a couple of months ago. Your #2, well, as I said I always thought the IFR for the general population was under 1% so the fact that it appears to be under 1% was already factored in. #3 seems like a minor variant of #1: if people need more vitamin D they can take a supplement or get more sun. Your #4 I did not know. #5 can go either way: on the one hand the novel coronavirus doesn’t kill young people the way the 1918 flu did, so that’s great; on the other hand it appears (from the examples of Italy, Spain, and New York) that it still has the capability to overwhelm our health care system and lead to a huge number of premature deaths, by a larger factor than the flu did (although not perhaps if measured in live-years-lost). Anyway it seems plenty bad.

      So..’optimism’, I dunno. I’m certainly optimistic that we can avoid a nationwide tragedy that is a scaled-up version of the worst of the local tragedies we’ve seen, so that’s great. But we still seem to have a choice between (utterly destroy the economy and kill a hundred thousand people, most of them old) or (mostly destroy the economy and kill three hundred thousand old people and 50,000 middle-aged people). Could be worse! It could always be worse! But this still looks pretty bad.

      • I definitely saw estimates, early- to mid-March, in the media of 2-3% fatality rates (I remember seeing 3.4% from WHO). These were CFRs, not IFRs, but the media didn’t make that distinction. I even saw claims that that made this deadlier than the 1918 Spanish flu at 2.5% (which is a baffling claim, given that 1918 H1N1 probably killed more than 2.5% of the *entire world population*).

        I didn’t entirely *believe* these estimates, because I figured there must be un-noticed mild cases (I remembered 2009-10), but the impact in 4/5 of the US so far is still definitely far less than what I was expecting mid-March. (I expected Florida to end up about as bad as New York, for example, since they were about equally late to act. I expected that a bunch of my co-workers would get sick with it, since we had a huge mass meeting just a couple of days before the closures started; it’s been 2 months and none of us have.)

        So ‘better’ is relative. Even with a degree of expectation that the media was probably erring toward the side of worst-case, what I’m seeing now (both locally, and in the US-wide numbers) is a lot less bad than what I would have expected March 13th.

        This might end up comparable to 1957 and 1968 pandemics, in per-capita terms (the deaths-per-million in 1957 would translate to about 210,000 in the current US population.

        I don’t mean this as any kind of actual prediction, of course, just that moving from “this will probably be about as bad as 1918, maybe somewhat less”, to “this will probably be about like 1957 and 1968, maybe a little worse”, is still a notably more optimistic outlook.

Leave a Reply to Phil Cancel reply

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