The Pandemic: how bad is it really?

This post is by Phil Price, not Andrew.

Andrew’s recent post about questionable death rate statistics about the pandemic has reminded me that I have not yet posted about a paper Troy Quast sent me. Quast is from the University of South Florida College of Public Health. Quast, Ross Andel, Sean Gregory, and Eric Storch have written a paper: “Years of Life Lost Associated with COVID-19 Deaths in the United States During the First Year of the Pandemic. [Note added later at Quast’s request: Here’s the published version in Journal of Public Health.]

Here’s the money quote: “We estimated roughly 3.9 million YLLs due to COVID-19 deaths, which corresponds to roughly 9.2 YLLs per death. We observed a large range across states in YLLs per 10,000 capita, with New York City at 298 and Vermont at 12.”

This is a follow-up of a similar paper from 2020 that appeared in the Journal of Public Health…similar in approach, although that paper covered only the first six or seven months of the pandemic. Quast very kindly credits me with inspiring this work via a post on this blog last May.  That post generated lots of interesting discussion, which some of you may want to revisit. 

[Yikes! I would have thought I could at least get the order of magnitude right when I divide in my head, but evidently not, or at least not reliably. As several early commenters noted, I bungled the numbers badly. The numbers below have been fixed, and the commentary to match.]

3.9 million Years of Life Lost (YLL), how does that help us gauge the magnitude of the pandemic? Well, we could compare this to  U.S. military losses in WWII, for example: that was about 400,000 people killed, at roughly 50 years each, so 20 million YLL. In the first year of the pandemic, the U.S. lost about 1/5 as many years to COVID as we lost in all of WWII, 

Daniel Lakeland has suggested that YLL is most easily interpreted if you divide by 80 to get ‘equivalent lifetimes’. 3.9 million years of life lost corresponds to about 50,000 lifetimes. 

In 2015 the National Transportation Safety Board published a table of leading causes of death, ordered by years of life lost. At 3.9 million YLL lost in a year, COVID would be third on that list (after cancer and heard disease), and more than double the next contender (chronic lower respiratory disease).

A bit of a tangent, but in case you’re wondering the 2015 list, in order by YLL, was: Cancer; Heart disease; Chronic lower respiratory disease; Accidental poisoning; Suicide; Stroke; Motor vehicle crashes; Diabetes; Chronic liver disease; Infant death. This shocks the hell out of me. I knew there are lots of suicides, but I would have bet traffic accidents were responsible for way more YLL than suicide, but no (it was quite close, though, at least in 2015. And I would never have put ‘accidental poisoning’ at number four, probably not even in the top 10.

But I digress. 

Check out the paper, there’s some pretty interesting stuff in there, such as: “In every jursidiction, the male value is greater than the female value, which is consistent with YLLs per 10,000 capita for males being roughly a third greater than for females at the national level (136.3 versus 102.3). However, the divergence between the two genders varies considerably by state. In New York City, the male value was nearly 75% greater than the female value, while in Mississippi the male value was only 7% greater.”

 

This post is by Phil 

70 thoughts on “The Pandemic: how bad is it really?

    • It depends on the level of granularity of codes they chose. The linked document says they used icd-10, but not a breakdown of codes. For poisoning the codes can get at both substance and intent. Poisoning for suicide should be given as suicide as reported here, poisoning to murder as murder, etc.

  1. I’m confused. How do you get 3 million YLLs for WWII? 400,000*50=20 million. More generally, the number of US lives lost is comparable between WWII and Covid-19, but the ages of those lost are, on average, quite different. In addition to that arithmetic problem, I think adjusting for the difference in population in the US would be appropriate. Put differently, per-capita YLL would be more meaningful. If per-capita YLL in the US and some small country are similar, the pandemic isn’t somehow much worse in the US simply on account of its greater population.

  2. Phil:

    Approx 500,000 deaths x approx 8 years of life lost per case = approx 4 million years of life lost. I guess this is consistent with most covid deaths being old people who had, demographically speaking, 8 years left of life to live?

    Also, yes, what Ian said. U.S. dead during WW2 is of the same order of magnitude as those killed by covid (so far), but the soldiers were young, so if you give each of them 50 more years of life, that would give total years of life lost to be something like 5 times more than covid.

  3. Phil –

    Seems to me that this concept only scratches the surface of measuring “how bad” the pandemic is. Not to say that YLL should just be ignored, but I’m not sure why is would be singled out as overall metric. How meaningful a metric given the full scale of how bad its been?

    Obviously, illness is an important impact of the pandemic. How many people have been hospitalized for every one that died? How many people will have long term sequelae? How many healthcare workers will be impacted long term from the psychological impact of caring for so many desperately ill people?

    How do we measure the disproporiinate impact in terms of morbidity and mortality in particular communities?

    How do we measure the economic impact? The impact of kids missing in-person schooling? The psychological damage from so much isolation?

    And as you might remember, I question the idea of measuring the relative impact of different deaths via a simplistic construct that the impact is directly proportional to the years of life lost. How do we meaningfully compare five years of potential life lost with one death to 50 years of potential life lost to another death?

    • This all got hashed out in great detail in that post last May.

      I don’t think anyone would argue that any single metric captures the effect of the pandemic. I certainly wouldn’t. And yet, the popular press focuses almost exclusively on ‘deaths’. It’s pretty unusual to see discussion of something else, such as the decrease in life expectancy or whatever.

      Me, I want to see the ‘deaths’ number but it’s so deficient that I definitely want to see something else too. We don’t have to pick just one metric, but if I did I would probably prefer YLL to deaths. You might feel differently.

      • I think this is a really great descriptive analysis, so long as we don’t use it to guide ‘how should we have reacted and what should we have done?’
        Channeling my inner Nassim Taleb a bit here: stochastic multiplicative processes have unique properties and should not be treated in the same category as additive risks like heart disease, falling off ladders, whatever. This seems like an annoying point but it is absolutely fundamental to framing what a truly rational response should look like! Discussions like these often just further arm the ‘Know Nothing’ crowd without that strong proviso

      • Phil –

        In response to a comment of yours from downstairs:

        > It’s funny that you think YLL gets “undue attention”; I pretty much never see it outside technical or quasi-technical discussions.

        I think that comment largely explains our different orientation on the utility of YLL. In the context of COVID, I’ve seen it use mostly as a rhetorical device by people who are critical of attempts to mitigate the pandemic, by the way of minimizing the significance of the impact of the pandemic.

        In response to your comments here:

        > And yet, the popular press focuses almost exclusively on ‘deaths’. It’s pretty unusual to see discussion of something else, such as the decrease in life expectancy or whatever.

        and

        > but it seems crazy to me to rely entirely on a metric that doesn’t distinguish between losing 1 year of life and losing 50.

        This is also somewhat different from my experience; while no doubt the count of deaths gets the bulk of the focus, I’ve also seen a lot of focus on many other impacts from the pandemic – including the extreme skew in age stratification of deaths and severe illness, as well as other demographic aspects of the impact from COVID.

        Maybe part of the difference in our experience is that I tend to look at rightwing media as a regular practice. Nut even outside of that media environment, I don’t see that people are “rely[ing] entirely” on that one metric.

        > We don’t have to pick just one metric, but if I did I would probably prefer YLL to deaths. You might feel differently.

        Yah. If I’m making choices, neither would EVER be considered as a stand-alone metric without consideration of other context. So I don’t see a reason to even speculate as to which is the better metric in some kind of dichotomy. I don’t see either metric as having some kind of an objective superiority; they’re just different perspectives on the same phenomena.

  4. Here’s a different way of comparing deaths from Covid with deaths from other causes. I’m wrapping the table in … tags, so here’s hoping it will be readable. The US covid death rate range was from a time last year near the peak death rates.

    The idea here is that in the US, we seem to accept the fatality rate from automobile crashes, and even from suicide. How much higher than these are the Covid death rates? They have been much higher. They are even higher than the most intense WW2 battles. Only the worst Civil War battles had a higher daily fatality rate, and those were only for a few days at most.

    (war numbers from

    http://www.bkdunn.com/2008/03/us-war-deaths-per-day-by-conflict-war-battle-and-how-iraq-compares/
    )

    Cause US Deaths / Day Scaled for
    2020 population
    —————————————————————————–
    Auto crash 82
    Flu (varies) ~68 – 137
    Homicide (2018) 52
    Suicide ~123
    WW2 combat (overall average) ~50
    Invasion of Normandy (WW2) 37 86
    Battle of Iwo Jima (WW2) 194 453
    Battle of the Bulge (WW2) 470 1097
    Bull Run/Manassas II 1000 10414
    Battle of Gettysburg 2621 28195
    Battle of Shiloh 1741 18130

    All Causes 7821
    Covid-19 1000 – 3100

    • Grrr, the formatting didn’t work. Sorry that the table is harder to read than intended. The first number in a row is the reported deaths per day, the second number is the number of daily deaths rescaled to the 2020 US population.

      For example, in the Battle of the Bulge, there were on average 470 deaths per day. Scaled to the 2020 US population, that would have been 1097 deaths per day. Still much less than for Covid.

      • Interesting but it’s not intuitive to me how something like a particular battle scales to modern times. In modern times we’d expect casualty rates to be much lower; in one way that has nothing to do with anything yet it’s the first thing that jumps into my mind.

        • Good point — doesn’t make sense to me either. In particular, medical treatment has presumably improved considerably since WW2, so that there would be a lower rate of death among the wounded.

          (As a side note: This example was particularly interesting to me because my oldest cousin was a fighter pilot for the US in WW2, and died when his plane went down in October, 1944, on a mission over what became East Germany after the war.)

  5. Pardon my naivete, but I don’t understand the conceptual basis for YLL as a summative metric. Is the essential human good that public health seeks to maximize the total number of years lived by the population? Are those years ends in themselves or is their value based on the activities people fill them with? If we had more people, for instance with a higher birth rate, would that be a positive addition to our sum of life years just as premature death is a subtraction? What do we think about two 70 year olds, each with a ten year expected future lifespan, dying of Covid relative to one 60 year old?

    Of course, YLL is an interesting metric insofar as it bears directly on life expectancy when denominated by the population in question. But life expectancy in only one of several meaningful measures of well being, and if there’s a theoretical basis for privileging it, I don’t see it yet. Life expectancy can be a useful proxy for other health metrics with which it tends to be correlated, but then it derives its extra importance from these other metrics.

    • I should add that I realize the authors of this article & others who estimate YLL’s don’t necessarily claim it is the be-all metric; they just think it’s interesting and have a go. I am reacting to the undo attention it sometimes gets. I admit I’m guilty of having used DALY’s in some prior work, but I think the effort to incorporate disability makes it at least a little more defensible. (Being skeptical of utility theory, I’m not a fan of QALY’s.)

    • Peter,
      Name a metric and I’ll tell you a bunch of things “wrong” with it as a way of summarizing the effect of anything that causes premature death. Just counting “deaths” implicitly counts each remaining year for a 90-year-old as if it’s 20 times as valuable as each remaining year for a 10-year old. That doesn’t seem right. You’ve already said you’re not a fan of QALYs, so presumably you recognize some things wrong with them. You don’t like YLL either; that’s fine. You got any suggestions?

      It’s funny that you think YLL gets “undue attention”; I pretty much never see it outside technical or quasi-technical discussions. The number I see most often (by far) in the context of the pandemic is “deaths.” I’ve got no real problem with citing the number of deaths — it’s very much less model-dependent, it’s easy to explain, every death in this country produces a death certificate so the counting is easy, and so on— but it seems crazy to me to rely entirely on a metric that doesn’t distinguish between losing 1 year of life and losing 50.

      You can pick your poison. If you don’t like YLL, you don’t have to pay attention to it! I’m with you on QALYs, I don’t pay attention to those myself.

      • So I think part of the surprise in suicide and accidental poisoning being higher than you would have guessed is probably not due to it being more common than you expected but the ages of the people doing the dying. And the improvements in vehicle safety (and in people abiding by the laws requiring their use.) in the last 20 years.

        • You’re right, this is surely part of it, especially for suicides.

          But there’s more going on, I think I just have some skewed ideas about how people die. Or…ah, it just occurred to me, I bet “accidental poisoning” includes drug and alcohol overdoses. Maybe that’s obvious to everyone else, but it didn’t occur to me. OK, yeah, I can believe lots of 16- to 30-year-olds are accidentally overdosing.

          I wish I could return myself to the state of ignorance I was in before I saw that table, so I could try guessing the order and the numbers and see how badly I’d do. Pretty badly.

      • I’m reacting to the role that YLL has played in various burden of disease studies and especially in monetary valuation, where the value of a YLL is the main alternative to plain vanilla VSL. (Cue Cass Sunstein.) I don’t like either very much. I suppose I was unclear in not specifying whose attention I was referring to; I wasn’t thinking of journalistic coverage of the pandemic, for instance.

        I completely agree that any single metric leaves out matters of importance; “health” is situated along many dimensions. And age at death is certainly important — it was one of the reasons AIDS was (and to an extent is) so devastating. But I would also be wary of any metric that leads us to just shrug our shoulders at widespread nursing home deaths. If popular accounts are accurate, there was some of this in Sweden, for instance.

        • Yeah, I too am appalled by some of the comments about deaths among old people. I think some of this may be less left-vs-right than it is based on people’s personal experience with old people. My dad was very ready to die at age 82 and his death (non-COVID-related) was welcome. If most of the old people I knew were like that, I, too, might shrug my shoulders at those deaths.

          But my 80-year-old mom is extremely healthy and happy and vital, and I hope has many more good years ahead of me. She and I have started biking over the hills to a smoothie place, getting snacks, and biking back…about 35 miles with 3200 feet of climbing. Sure, she’s on an e-bike, but it’s still a bike, not a motorcycle, and she has to work hard to make it go up those hills and to save the battery enough to finish the ride. Since I know several people her age or older for whom the remaining years are still bright, I understand the tragedy of a death among the old.

          That said, though, I haven’t seen people quote YLL as a way to minimize or understate the effect of the pandemic. I -have- seen people say something like “yes, a lot of people died, but they were mostly old people” and other NON-quantitative statements like that, that I believe are intended to understate the tragedy. To me, one of the interesting things about the YLL number is how big it is, per death. Someone can be pretty old and still have quite a few years of expected life ahead of them!

        • >> I think some of this may be less left-vs-right than it is based on people’s personal experience with old people

          Yeah. I feel like if I were, say, 80 and just diagnosed with something neurodegenerative and incurable like Alzheimer’s, I wouldn’t have a very strong motivation to avoid things like COVID. (That kind of mental decline is actually more frightening to me than death.)

          But I know/have known people well past 80 who are very happy, active, and involved in the world (a company I used to work for had an 85-year-old guy come in and do repairs!)

    • Peter –

      > Life expectancy can be a useful proxy for other health metrics with which it tends to be correlated, but then it derives its extra importance from these other metrics.

      As I think about it, sometimes I feel that each of the years I have remaining have an increasing value as I get older. Certainly there are cultural frames were “elders” have a particularly high value to society.

      • I think this can go different directions, depending on the person. My dad welcomed death because he was aware of his cognitive decline and mental health issues, and dreaded what the future years would bring. My mom, at the same age, is (so far) full of life and energy and has suffered no noticeable cognitive decline at all. My dad considered additional years of his life to have negative utility; my mom is still very strongly positive.

    • For just a rough comparison of the scale of the pandemic YLL seems fine.

      But I’m surprised that no one has yet mentioned – or maybe not! – that the long term **economic** impact depends strongly on the age of the people that are lost. It’s unseemly to say so, but the loss of people over 70 will not have a significant long-term impact economically, since these people contribute relatively little economic productivity and they are mostly not even contributing their experience to younger people. Their loss may even be a benefit over the long run as the government could save billions in social security and medical expenses.

      I have read that a key destructive aspect of the AIDs epidemic in Africa is that AIDS frequently kills middle-aged people in the prime of their lives and careers, thus depriving the economy of their high level of productivity and experience. This will certainly **NOT** be the case for COVID, and it might be one reason the economy is bouncing back quickly.

      From an economic standpoint a good metric would be YLL/person. I’d suspect that YLL/P < 15 would be relatively benign economically, vs. YLL/P in the 25-45 range would be highly destructive economically.

  6. Phil, why does this metric tell us “how bad the pandemic is, really”?

    To put that differently: In your opinion, what truth is conveyed by this measure that other measures misrepresent?

    • It’s not that the most common metric, “deaths”, misrepresents anything; it’s just really deficient, I think. As I said elsewhere on this thread, it seems crazy to me to rely entirely on a metric that doesn’t distinguish between losing 1 year of life and losing 50.

      • Hold on, you’re not using a different metric, you’re just weighting it differently.

        A different metric would be, say, total years of life lost, which would take into account years lost for people who caught it had to didn’t die.

        What you’re using is still deaths — that’s the trigger that gets individuals included in the count — weighted by how long they would be expected to live otherwise.

        On other words, it’s deaths but infants are vweighted at, say 25 times the very old, and about twice the middle aged.

        So again, why does this tell us how bad the pandemic is **”really”**? As opposed to unweighted deaths, which your headline implies tells a story — but not the “real” story.

        Personally, I think it’s crazy to weight the life of an infant at double that a middle aged person, but I’m sure you have your reasons. I’d like to hear it.

        • As I’ve said on this blog long ago, I think the biggest loss is someone in their late teens or early twenties. All that time and effort bringing them up to speed, and then just as they become whole and productive they die.

          I think “deaths” is not a good metric when the age distribution is extremely skewed, as it is with COVID. No single metric can capture the badness of the pandemic but this does not mean all metrics are equally bad. I think ‘deaths’ is a very bad metric and YLL is a lot better. If you want to see more discussion of this by me and others then please look at the blog post from last spring (linked in the post above), I’m not going to repeat it all here.

  7. 300 YLL’s /10000 people
    .003 YLLs per person
    10 days lost per person
    I just can’t argue a one time loss of ten days per person is that bad. Also the comparison to Ww2 is in absolute terms, which is misleading because the rest of the article is in per 100,000. Also just doing per person is easier because we are people.

    • Tim:

      I’m confused by your calculation. Phil said 3.9 million years of life lost. The U.S. has 330 million people, so that’s 0.012 years of life lost per person. 0.012 years is 4.3 days, not 10 days. 4 days is not a lot, so I guess that in a theoretical world in which people could just accept the deaths, it wouldn’t be such a big deal? It’s hard to say, because people react very strongly to the risk of getting exposed. Business as usual isn’t a realistic option. And if everybody had done business as usual, we’d’ve seen more coronavirus deaths.

      I guess we could also compare to other hazards. About 40,000 people die each year in car crashes. If we guess that on average they lost 40 years of life, that would be 1.6 million years of life lost, hence 1.8 days of life lost on average. That’s pretty small, and indeed people get in their cars every day without worrying about it. On the other hand, we as a society spend a ton of money on car safety features, road safety, cops, ambulances, etc.; I guess that’s the analogy to all the coronavirus precautions.

      WW2 was on the order of 20 million years of life lost, out of a U.S. population of 135 million. That’s 0.15 years of life lost per person, or 54 days. I guess that losing 54 days of life on average was worth it, considering the alternative?

    • That’s a really interesting way of framing it…one that seems really odd to me, but perhaps that’s because I’ve never thought about it this way.

      First, by this standard pretty much nothing is worth more than a shrug. Heart disease is the leading cause of death in the US and is responsible for about 8 million YLL per year, that’s only twice as many years as we lost in the first year of the pandemic. If the pandemic isn’t that bad, then the leading cause of death in the country is also not that bad! Cancer, accidents, suicide, homicide…not a big deal. Don’t worry, be happy!

      It’s an interesting perspective, that’s for sure.

    • How does a population average tell us anything meaningful?

      If you and your entire family were murdered in the capitol insurgency, should we all just shrug our shoulders because, on average, that’s less than a minute of expected future lifetime per person in the world? I mean really, I waste more time than that each day deciding what to have for breakfast. Who cares?

      • I’m not saying I agree with this – but I think it is *possibly* relevant here (unlike most other causes of death) since the measures taken to avoid/reduce COVID deaths had significant impacts on the majority of the population.

        It’s not clear to me that the average American lost less “expected quality of life” from staying home etc. than they would have lost by getting COVID.

        Now, this isn’t really a fair comparison, because the level of fear was high enough that normal life IMO wasn’t possible.

        Arguably, in a hypothetical world where everyone thought in terms of statistical risk assessment — or maybe even in a world where the media climate and risk tolerance was more like it was for the 1957 and 1968 flu pandemics — we would have been “better off” accepting somewhat higher levels of COVID in return for less disruption, but that’s not the real world.

  8. So they write “In our analysis we conservatively reduced the expected life expectancy by 25% to reflect the typically
    greater morbidity of COVID-19 decedents” which is really not conservative at all. I’m not even sure reducing it by 75% is “conservative”.

    • Yeah, I’d like to see more discussions of this. 10%, 20%, 30%, how do we choose? I think one can get some decent estimates by looking at excess deaths by age group (compared to typical years).

      My gut feeling is that 25% is indeed a bit conservative and 75% seems way too high, but I could be wrong.

      • Phil:

        Part of the difficulty is that “conservative” has no clear meaning here. It could be considered conservative to do a very small adjustment, or it could be considered conservative to do a very large adjustment. I think you and Matty are implicitly using different definitions of “conservative” here.

        • I think in this case that is not the issue.

          I think Matty is saying that the YLL is much less than the estimate in the paper. The paper tries to be “conservative” in the sense that their estimate is more likely to be an underestimate than an overestimate of YLL. The paper attempts to be ‘conservative’ in this sense by reducing the assumed life expectancy of COVID victims (i.e. how many more years would they have lived if they didn’t get COVID).

          I think Matty believes most of the COVID deaths are among people who would have died much earlier than the paper assumes (presumably due to co-morbidities.) That seems possible to me, but unlikely. Yes, people who die of COVID are, on average, less healthy in other ways than those who don’t, but lots of people who are not unhealthy compared to the average person in their cohort have died of COVID. How many? I don’t know precisely, but it wouldn’t take that many for Matty’s “75%” number to be untenable.

          My gut feeling is that the assumption in the paper is indeed somewhat more likely to underestimate than overestimate the YLL, but, as I said, I’d like to see the paper go into this more.

        • Hmm, yeah, this is the big question.

          I would kind of expect the adjustment to be pretty large, since even at very old ages many more people survive than die, and I would think that who dies of COVID and who doesn’t is far from random (ie largely based on initial health).

          Now, there are random factors in, say, the immune system producing antibodies, initial viral dose, etc., but still…

          >> I think one can get some decent estimates by looking at excess deaths by age group (compared to typical years).

          Not yet, I think. In 5 years, yes. For this purpose, I think the question is *how much* those deaths were sped up by. If US deaths are *less* than usual in 2022 (because a lot of people “scheduled” to die then died in 2020-21) that’s a different picture than if they are not.

    • This is a fundamental issue with this analysis. What’s even the counter factual here?

      If the counter factual is YLL for non-COVID cause — if I hadn’t gotten COVID how long would I have lived — then by definition heart disease etc. cause zero YLL lost and so COVID is infinitely worse which is clearly nonsensical.

      • I don’t follow. You always compare the age at death from people with condition X to the age at death for people without condition X. If X is heart disease, you compare people who died of heart disease to people without heart disease. If X is COVID, you compare people who died of COVID to people without COVID. And so on. For any condition that is fatal in some people, you will get a non-zero answer. I’m not sure what you’re imagining, but if it leads to an estimate of 0 YLL for heart disease then you’re imagining the wrong thing.

        • At some point everyone dies of “something”. if you die of something that is almost exclusively the killer of extreme elderly people, it’ll look like that thing is “good for you”.

        • Sorry, let me sketch this more carefully as I think we’re mostly saying the same thing.

          First: for covid specifically there’s a clear treatment and counterfactual — let’s say a magic pill curing covid — and we can define the YLL via the treatment effect of that magic pill w/ endpoint death. Call this T1. Of course measuring the effect of T1 is impossible; the paper estimates this by assuming 75% of the E(years left | age). Is this estimate correct? Who knows, but the entire paper hinges on the choice of arbitrary 0.75.

          Second: for (say) heart disease, there is a (perhaps) distinct population with treatment T2 — say magical heart disease cure — and corresponding treatment effect. We can again estimate the effect and get a number. Of course ordinally you can compare the two effects, but I question whether this comparison is meaningful.

          For ex. let’s assume the effect of T1 on the heart disease population is zero. OTOH, the effect of T2 on the covid population is very likely not zero! Probably a magic cure for heart disease would be impactful for covid as well. Since T2 is more impactful, you could then conclude heart disease is worse… but mainly my point is this is the wrong setup for comparison.

          A more meaningful setup would consider a single population, some of whom have heart disease and/or covid, and two treatments T1 and T2 along with the single endpoint of death. Who knows what we’d find, but very clearly “which is worse” in this setup depends crucially on the copula.

    • Yeah, I’ve wondered about this. 8 or 9 years of life lost / COVID death doesn’t sound unreasonable, but OTOH I’ve seen *really* short life expectancies given for nursing home residents, which (until about 2 months ago when vaccination really kicked in) made up a very disproportionate number of COVID deaths.

      And one would kind of expect that even among a very-high-risk population like that, those with the shortest remaining life expectancies (ie least healthy) would be most likely to die.

    • We know the main risk factors for Covid mortality are age and cardiovascular health. The model they use properly accounts for age, which is easy. The they try to account for cardiovascular health and all the other smaller risk factors by throwing in a 25% reduction in the presumed life expectancy of those who die.

      Given that someone who is 79 (the median age of Covid decedents) has a life expectancy of about 7 years, reducing that by 25% gives 5.25 years. Yet, only about 5% of people that age who contract Covid die from it. Now, we cannot assume that those who die perfectly overlap with the least healthy 79 year olds, since many will die of causes that do not impact their risk of death from Covid. But, it seems reasonable to suppose that the decedents more nearly overlap with the least healthy than that they are representative of the bottom 75% of their age group. The study essentially says that the 5% who die from Covid have the same pre-Covid life expectancy of the bottom 75% of their age group. This requires that the decedents should be a random selection from that bottom 75%. I find that highly unconvincing.

      Their methodology defies common sense. They don’t even present their 25% guess with a confidence interval; it’s just an arbitrary guess. Here’s my guess: Politics once again. Leftist politics, of course.

      • Craken –

        > Yet, only about 5% of people that age who contract Covid die from it.

        Hmmm. Well perhaps not directly relevant to your point about the article, but the latest CDC estimate is 9% for 65+, and I would guess that implies quite a bit higher than 9% for people aged 79.

        > Politics once again. Leftist politics, of course.

        I’m thinking by your logic combined with your assessment of the IFR for people aged 79, you must be a leftist?

        https://reason.com/2021/04/02/new-cdc-estimates-suggest-covid-19-is-deadlier-than-the-agency-previously-thought/

        • We’re in Month 15 of the pandemic (here in USA) and we’re still bandying about terms like “IFR” when we still don’t actually have any good information on, you know, how many people are infected (that’s the “I” in “IFR”). And even on a blog of, by and for stat geeks that seems to be shrugged off as not a serious impediment to drawing conclusions.

        • nwbr –

          I agree that there seems to be a lot of uncertainty. I think that uncertainty is an impediment to drawing (confident) conclusions.

          That said, I think that if you’re going to ballpark it, the CDC estimate may be a good way to go.

          But my larger point was to question Craken’s reverse engineering of causality, attributing a dubious article to “leftist politics.” I don’t think that such errors are likely proportionately distributed in association with political ideology. In fact, I think there’s little evidence supporting such a mechanism of distribution (although strength of political identification is likely associated with prevalence of such types of errors).

        • Considering that slightly less than 1% of the 75-84 age group has died from Covid, I think 5% is a solid estimate for the IFR of 79 year olds. To find something much higher presupposes a much lower infection rate among that age group than among adults overall. The overall adult infection rate is about 25%. If the IFR of 79 year olds were over 9%, as you propose, their infection rate would be below 11%. I just looked at the CDC report referenced in your Reason article. It specifies an IFR of 1.4% at age 65, 4.6% for people at age 75, and 15% at 85. This allows us to infer a 7-7.5% IFR at 79.
          Assuming the CDC is accurate, this does little to vitiate my criticism of yet another poor study. You failed to address my main criticism concerning the study’s estimate of the Years of Life Lost.

        • Cracken –

          > This allows us to infer a 7-7.5% IFR at 79.

          Maybe your inference is correct (I will own up to placing a higher probability in the CDC being correct than I do to your estimate being correct, at least without any better sense of the comprehensiveness of your analytical process) – without more data it’s hard to say. And I made an error in saying “higher than 9% for age 79,” when I should have said higher than 5% for age 79. And given the uncertainty involved, 7-7.5% for age 9 doesn’t seem at all unreasonable.

          >Assuming the CDC is accurate, this does little to vitiate my criticism of yet another poor study. You failed to address my main criticism concerning the study’s estimate of the Years of Life Lost.

          Yah. I wasn’t commenting to vitiate your criticism of the article. My point in commenting to vitiate what seemed like the faulty logic whereby you reverse engineered from the errors in the article to the “of course” part of the “Leftist politics, of course.”

          Are you going to walk that part back?

      • I don’t think it’s so much political as the unfortunately common “we don’t know how large this effect is, so it must not be that important”.

        • I think the way it becomes political is that when everyone (including people who by training and career experience ought to know much better) get in the habit of “we don’t know how large this effect is, so it must not be that important” kind of thinking. If nobody is going to apply any discipline to that kind of thinking then the decisions about what should and should not be disregarded becoming, well, political.

        • 25% is a pretty substantial haircut, not really consistent with thinking “it must not be that important.”

          How big would it have to be in order to be “important”? 50%?

          25% seems reasonable to me. So does 20%. So does 30%. 50% seems like stretching it, and 75% seems extreme, I’d bet against it. Why do I feel this? I’m not sure, and I admit I could be very wrong.

          I note that countries where most people have far fewer comorbidities than we have here have still had high death rates. For instance, Sweden has had 1400 deaths per million people, compared to our 1770. It’s my impression that they have a much healthier and fitter population than we do, and they certainly have a better public health care system, which as far as I know never approached overload like we had in New York and a few other places. It seems like the most parsimonious explanation is that even relatively healthy old people are at considerable risk of dying from COVID. This is not the only possible explanation.

          One of the reasons people think otherwise-healthy seniors are unlikely to die of COVID is that healthy seniors actually are unlikely to die of COVID. The CDC reports that 94% of COVID deaths were in people who had ‘contributing conditions’, with the top ones being : Influenza and pneumonia, respiratory failure, hypertensive disease, diabetes, cardiac arrest, and heart or renal failure, and vascular and unspecified dementia. https://www.jems.com/coronavirus/cdc-report-underlying-conditions-94-percent-covid-19-deaths/ This makes it seem like COVID is selectively harvesting the ill.

          The thing is, otherwise-healthy seniors are hard to find. https://vitalrecord.tamhsc.edu/10-common-elderly-health-issues/ says “about 92 percent of seniors have at least one chronic disease and 77 percent have at least two.”

          That is, if you kill a random selection of old people, you’re going to find that something like 92% of them have a ‘chronic disease.’ So finding that 94% of COVID deaths are in people who have a ‘contributing condition’ is doesn’t actually tell us that COVID is selecting the sickest of the sick seniors. It could be randomly striking old people, as far as these stats go.

          Or, rather, this is _roughly_ true. I’m 100% sure the pattern of COVID deaths does NOT look random among a group of any age. I’m just saying the impression that only the sickest of the old people are being killed by COVID is not necessarily implied by the fact that almost all COVID victims have an underlying condition: almost all old people have an underlying condition!

          I do wish the authors of the paper had explained why they think 25% is conservative. But it’s not obvious to me that they’re wrong about the claim.

        • I don’t know what’s right, and you’re right that 25% isn’t that small an effect in absolute terms, but…

          >> I’m just saying the impression that only the sickest of the old people are being killed by COVID is not necessarily implied by the fact that almost all COVID victims have an underlying condition

          … this is true as far as it goes. But a lot of these conditions aren’t “binary” but have a range of severity. So it would seem odd if slight, well-controlled hypertension/heart disease/diabetes increased your risk vs. not having it, but severe/poorly-controlled didn’t increase your risk vs. slight/well-controlled.

          IE even among “people over age X with condition Y” there should be a range of severity.

        • I think they do try to explain why their estimate is conservative, and I read the cited paper, but didn’t follow up on the math because I had to do other work.

          They looked at a paper (that they cited right before their estimate) that had done some modeling on covid deaths in stratified age groups. And at covid deaths in people with co-morbidities. And at YLL in people with comorbidities, stratified by age group (unrelated to covid, because they were from long term health records from the last decade). Combined the YLL due to comorbidities, with the prevalence of each within the age groups and adjusted the YLL lost due to covid down to account for reduced life expectancy of the majority individuals dying from covid. I didn’t do the math on their average values and how that relates to 25%. Maybe if I did the math or read the other paper more completely I would have a better idea how conservative their numbers are. But I guess one can say they are more conservative than not accounting at all for the effect of comorbidities.

          I did do a little math.

          And like you mentioned most of the older people already had some of those conditions, And reducing the YLL for someone who is 80 from 10, to 7.5 seems reasonable especially given that the Actuarial tables that the life expectancies are estimated from already incorporate the presence of morbidities for seniors.

          The worst pre-existing condition for someone who isn’t a “senior” is Obesity (I know this from published data and also from my husband treating hundreds of covid patients in ICU over the last 14 months). And reducing the life expectancy of an obese 45 year old from about 80 to about 70 seems reasonable also.

        • Hey, no fair bringing actual data and analysis into it, we were having a great time speculating and sharing anecdotal ‘evidence.’

          Of course you’re right that if we’re going to discuss the validity of that 25% number, we should at least glance at the paper that informed it. The rest of us should feel a bit embarrassed here. I know I do!

Leave a Reply to jim Cancel reply

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