Warning

Warning: This site contains images and graphic descriptions of extreme violence and/or its effects. It's not as bad as it could be, but is meant to be shocking. Readers should be 18+ or a mature 17 or so. There is also some foul language occasionally, and potential for general upsetting of comforting conventional wisdom. Please view with discretion.

Monday, October 5, 2020

Covid 19 Deaths, Not Lockdown Deaths, Part 1: Why the Deadly First Month?

Covid 19 Deaths, Not Lockdown Deaths, Part 1: Why the Deadly First Month?

October 5, 2020

Lockdown Kills! We've heard this around from people who think the SARS CoV-2 virus itself only kills on par with a seasonal flu. Yet all over, a serious outbreak leads to alarming daily deaths from confirmed infections, and then an enormous number of people past expected norms die on top of that, far above what the virus is known to kill, but piling up just when its deaths do - primarily in the 3-4 weeks after lockdown orders are imposed. 

Alarm bells!

What started as a response to just two of the many articles around making such claims turned into a strong but expansive general rebuttal that needs to be done in three parts. I'll start with virus apologist John Pospichal's "Questions for lockdown apologists," which Vanessa Beeley for one passed on as a "very well researched article." He clearly put some work into it, and it's astounding how he was able to follow through that far while completely missing the fairly obvious answer that whole time.

Peak Deaths AFTER Lockdown - Why?

After a quick read, it seems Pospichal just asked variations of one (compound) question: why did Covid 19 deaths, and indeed overall (all-causes) deaths, spike suddenly after lockdowns were imposed, and remain high for weeks afterwards, and not rise noticeably before that, despite the virus apparently spreading for months already? The suggestion is that the severe restrictions are what killed so many over the first few weeks. Just how and why isn't made clear, but the patterns were supposed to speak for themselves in showing it must be something else.

The discussion of early spread is interesting, seems true, and it's worth asking why it didn't kill more widely before (except it seems in Spain, S. Africa, somewhere in the U.S. by Christmas, anywhere a below-average baseline hides the small early death toll, etc.)  Many, many places see unexplained surges of deaths in mid-February, some building on one already underway. But these surges tend to drop - apparently on their own, or from quiet control measures - before the covid surge noticeably begins, variously in mid-February to mid-March. 

Perhaps a more contagious strain started spreading in January, or it just started spreading from far more points in quick order, adding to the existing transmission pathways, hitting the point of exponential growth where it goes from all-but invisible to overwhelming in very short order.

Whatever to make of that issue, once the new wave of infections took off, late February up to lockdown, it got bad all over quite fast. As cases and impending deaths rose sharply, escalating measures were taken. And yet the impending deaths rose sharply for a couple weeks after the final lockdown orders, as if the orders did NOT magically cure them. 

Virus and overall deaths continue past the peak because ... most of them are virus deaths, and in those death comes well after infection and confirmation, not the same or next day. In fact it can take just a few days up to a month, and usually 2-3 weeks, with several sources agreeing on 18.5 days median time between first symptoms and death. For example, Drugs.com gives 18.5, while indicating 19, in a chart I adapted below to get a better view relative to actual infection date.


I'm still no expert, so keeping it broad ... 2-14 days is commonly given as the broad span between infection and symptoms. Confirmation date will vary a lot too, from first concern to first symptoms, to several days after that to ICU admission around day 12 (to death, to never). But confirmation doesn't matter here. Broadly, Covid 19 deaths will follow infections by a wide range of between 20 and 33 days. That may be a bit late and long, and it will describe just the majority of cases involving a real struggle. There are some with bad prior conditions where just a few days of infection is enough to push them over the edge, so a number of deaths resulting from any one day's infections will come scattered over the days and weeks before the crowd arrives at the final finish line.

So if you look at the day after lockdown, you'll see deaths already set in motion about a month back. If infections are to fall off, check for that around 20 to 33 days later. If the response was good, I reason you'll see them already falling by then.  

Working in this key detail to the handy graphic Pospichal had prepared flips his case around nicely. Deaths peak usually 15-20 days after lockdown, maybe depending on how well prior measures had worked. They always fall steadily during the 13-day span roughly corresponding with the last pre-lockdown fatal infections. new cases/deaths will have come in, and a new surge might be waiting in the wings if measures failed, but here we see the high death tolls Pospichal wanted to claim as lockdown deaths all show how well lockdown saved lives.

Madrid: 

Ile-de-France aka Paris, I think. A bit of a slow response, but then quite effective, as seen from 21 to 33 days out and beyond. Good thing too - notive it peaks at nearly 5,000 deaths in a week - the norm it had been was just over 1,500/week. Anyone care to see what would happen if they did lockdown a week later?

London: unusually, that plate had its control measures marked in red a week earlier than they should be, putting more deaths after the orders and less of them before. Interesting. Corrected here, checking 20-33 days out, something deadly was in fact locked down pretty well, besides the people.


New York City, a bit more inexact timeline here at right, but it seems to be the same picture: Lockdown worked, completing improvements made at least a week earlier with other measures (the slope goes from almost vertical to 60 degrees before peak).


Finally, hard-hit Guayas province, Ecuador (below). They saw peak deaths near 700/day 18 days after lockdown, with a sharp fall starting day 19, down to 200/day by day 33, and to average levels again by day 37. That seems to have worked out exceptionally well.



Conclusion

Everywhere, after 18-21 days plus some resolution, these last pre-lockdown cases have died, and deaths then fall steadily by day 33 to fairly low levels, taking different turns from there. That means the lockdown orders and preceding measures worked. The main question on the death delay is answered.  Lockdown did not cause those ongoing coronavirus deaths - it stopped them from rising higher over a longer time. 

To note: the drastic shutdowns and stay-at-home orders so widely used did not do this single-handedly. Social distancing and special protection of the most vulnerable, by themselves and others, seems to have an even bigger impact on falling death rates. These can be seen lowering deaths prior to full lockdown, and must be behind the consistently lower fatality among older people during secondary outbreaks over the summer and fall. But the spikes of death we do so still see - smaller and delayed - show how they can avoid the virus much better when it's not being spread like mad. 

And we know lockdown didn't do this cheaply; the disruptions have been enormous. But the global surges of deaths this year are not one of these costs.

But if one still insists it really wasn't the virus but some aspects of the response, or anything else that killed these people, well... now we can add a detail, in case anyone wants to find out just what is truly to blame. The mystery cause(s) overwhelmingly kill on just about the timescale Covid19 would, peaking about 20 days after lockdown and falling from there. There are some other details we can add, and will in part 2 fairly soon.

Saturday, October 3, 2020

England and Covid19's Younger Victims Everywhere

 October 3, 2020

While the United Kingdom has suffered perhaps the worst death toll in Europe during the Covid19 pandemic, only the larger half of above-average deaths have been linked to the SARS CoV-2 virus. The rest are disputed excess deaths likely caused by unknown infections with the same rampant pathogen or, to many minds, by the lockdown itself. 

Among these, Kit Knightly at the Off-Guardian blog noticed, at the end of July, a trend in the mortality data. Among hard-hit nations, the vast bulk of victims were over 65, but England was different, seeing an "Unprecedented spike in deaths of 15-44 year-olds." Knightly found these, by age, "unlikely to be due to Covid19" and claimed - falsely as we'll see - "no (other) countries anywhere show increased mortality in people under 45." This "unprecedented spike," he says, "seems to be unique around the world. Why?" He offers no guess, except to doubt that it was caused by the virus, and suggesting the actual cause might be what I call "lockdown, somehow." 

The unusual level of deaths referred to is real, and we'll look at it in a little detail below. In fact I can add the issue has become a bit more pronounced since his article. Knightly takes 15-44 as too young to include many covid deaths, but this is sloppy thinking. As I'll show, it doesn't seem there that many, really. And also, 15-44 is broad, including some overlap with semi-old groups where fatal conditions apply more than one might think. Some quick basic references to start getting a view what rates apply around age 44, the upper end of the range in question:

compared to 18-29year-olds:
* 40-49 years had 3x higher hospitalizations, 10x higher deaths.
* 50-64 years had 4x higher hospitalizations, 30x higher deaths
* ages 35-44:  1,780 deaths, 1.722% of total
* ages 45-54: 4,976 deaths, 4.815% of total
* ages 18-39 had a  0.06 "hazard index"
* age 40-49 was 0.3
* 50-59 was 1.0, etc. 
By June 14 (citing old numbers since revised down): of a total 44,869 deaths: 40,023 were 65+, 4,359 were 45-64, and 487 were under 44. That small portion will be spread out over all the ages, but we should get why they'll mainly be clustered at the older end, with probably over half of them aged 35-44, and maybe even with over half of them in their 40s (and also predominantly male).

I'm still no expert, but when I looked into it early to assess my own risk, it seemed ages up to 30 had negligible deaths, with a slight increase in risk accumulating between ages 30 and 44, a slight but marked increase, it seemed, right at 45 (the line I'm just past). Of course it's down more to conditions, and I'm on the wrong side of that line too (maybe related to my anti-virus bias?) I don't recall now just where I had seen that and why it seemed so year-specific when most sources I see now just give ranges. But it had seemed to me the biggest single age difference was between 44 and 45, with the difference being smaller between 46 and each year up to around 55, where it gets increasingly worse every step from there up - I didn't pay as much attention past my zone, which we're dealing with here. 

If 45 is such a cut-off, a group spanning up to 44 might seem safely in the low fatalities zone. But in reality and on a big scale, there can't really that much uniform difference between the two ages. If 45 tends to be the start of high risk, 44 1/2, 44, maybe even 43 1/2 should see some early risk-bloomers. It would mean a sharper increase right around that age, over the slight increases seen from age 30-40. Reasonably, that means of the 487, likely 1/3 to 1/2 of this small number were aged 40-44. That would be 162 to 243 deaths = 0.36% to 0.54% of the total killed in the UK's first wave. There are already the scattered few in their 30s and early 40s this adds to. The total might be significant, if still paling in comparison to the masses of fatalities in higher age brackets. 

This pattern was noticed in combined mortality - covid and supposedly non-covid deaths. But the latter can in fact include any proportion of undiagnosed virus-related deaths, depending on the details (as well as vice-versa, also depending). My opinion remains a solid majority of those disputed excess deaths in the UK, and even some of those below average, were killed by SARS CoV2, following on confirmed, unconfirmed, and totally unknown infections. It kills almost always with the help of existing conditions, which it turs into co-morbidities, during or after active infection, by direct suffocation or exacerbated damage, causing death often decades ahead of schedule.  The relative few deaths in England and elsewhere under age 45 would be primarily among the 30-44 age sub-group, centered old (mostly 40-44). As far as I know, that exactly the age break down for all the elevated 15-44 deaths. 

I'll try to look out for, and will suggest someone should make available, relevant fatality breakdowns (esp. for the UK/England) for that overly-broad age bracket 15-45. But for now ... 

The stats cited above for UK deaths - again this is with the first wave up to late May/June, while the issue in question was only noticed in late July, in part because it was becoming more evident by then. If they were more covid deaths, they would follow the same age patterns, and that's just what's in dispute here; they seem to be too young. That pattern does exist and, as I'll show, it's gotten even clearer since then. England's "non-covid" deaths aren't just unusually young on average, they're getting younger all the time. And yet, the bulk of those above average (and even some of those below it) probably are more-yet victims of the same deadly virus.

Europe's Younger Death's Assessed

I checked deeper into the data Knightly cited (see below - the European Mortality Monitoring Project (EUROMOMO) - 20 European nations signed on to pool mortality data from 2016 on to help track public health threats like this one - there are signs of the same thing happening in other countries, most notably in Spain. Actual levels barely avoid reaching excess, but they do so after rising from the early low, sometimes for a couple weeks in a row, and at the right times to caused buy the virus. 

See here for the data cited: https://www.euromomo.eu/graphs-and-maps. For reference, the involved countries/locales are: Austria, Belgium, Denmark, Estonia, Finland, France, Germany (actually 2 entries: Belrin, Hesse - the rest of the country?), Greece, Hungary, Ireland, Italy, Luxembourg, Malta, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, and the UK as its four countries: England, N. Ireland, Scotland, Wales.

The Z-score they use for measurements will be explained somewhere ... the provided baseline uses historical weekly death tolls and some seasonal and yearly variables like heat waves into an algorithm that should decently predict the number of deaths to expect - basically - if this pandemic hadn't happened. (or so I gather on a skim of the site)

While the scale of it has no clear precedent among the examined examples, the issue of elevated younger deaths is not unique to England. How it compares to trends in, say Brazil, India, or Iran remains unclear, and even among the other EUROMOMO  states, we can see it's not entirely unique or unprecedented. In case after case, with varying clarity, we can see rises in deaths here corresponding to those among older classes and with known surges of coronavirus infections/deaths. 

And as I'll show, the rises in younger deaths are followed by larger surges among older folks aged 45-64 and more so with those 65+, just like covid death waves might move. The pattern was terrible in the pandemic's first wave prior to general lockdown and related control efforts, but greatly muted in the second waves of summer and fall. It's as if the older section were learning and avoiding the virus better while those aged 15-44 seem to keep getting it and then dying at about the same rate they always did. It's only natural then that the deaths are getting younger on average. And this would support a connection between all age groups -  same virus, different learning curves.

Upon a bit of examination, several countries' data is clear that parts of 2020 have been quite deadly for Europeans aged 15-44. Those parts tend to match up with Covid19 1st and 2nd waves, and otherwise might reflect less obvious levels of the same thing in between. There will be other causes mixed in bringing the death higher or lower than the virus alone would, but to imagine anything else causing the bulk of them ... is something I'd need a little help with.

Knightly picked for consideration the five European countries "most affected by Covid19 so far," which he fairly considered to be Italy, Spain, Belgium France and the UK. For each one, he showed a graph of weekly mortality for those aged between 15 and 44, squished small to show several years. England definitely stands out,  but at least Spain is visible even there as scoring significant excess at relevant times. Most of the others show different kinds of rises to and past the upper limit of normal, often after a quick rise from a level well below average, often corresponding to periods of strict lockdown (see France for example). 

Increases in younger deaths can be seen among all EUROMOMO nations combined: definite excess in this group past the red upper 5 at w12-16, and close to it w31/32, with two spikes to the upper 2 at year's start and February (possibly related), and a bit past that line in week 22 following the April peak, and close again week 37, with the most recent weeks elevated and holding steady just below the 2. In all countries combined, deaths 15-44 are well over baseline in the second half of the year, after starting a bit low. England will be the biggest contributor to this, but not the only one. 


Here I added the green box and other labels, including to explain the graph used throughout. The original chart also gives actual numbers of deaths, unlike most of them. Weekly deaths as high as 1,448 in week 15, to an expected baseline of about 1,170 - an additional 24% died across these 22 nations. It hits 14% over in weeks 31 and 32. )

Also note in the plotting above and those below a possibly misleading optical illusion: in most cases the actual number of deaths age 15-44 is unclear, and may seem large by the graphing. But usually this bracket dies off in fairly small numbers, with proportionally small variations that - by the nature of small numbers - look fairly dramatic and erratic. So it's not that hard to exceed normal or even excess, as soon as there's anything around to push them higher at all. If I read it right, the above plotting uses a baseline of around 1,150 to 1,200 deaths per week between all 20 involved nations. That 24% excess would only be an extra 288 deaths, across 20 nations and in a whole week.

So even England's massive spike looking similar to those for older folks still represents a fairly small portion of deaths. As noted above, only 487 Covid19 deaths under 44, plus whatever excess past that (another 300?)  fill out that massive peak up to June. A few hundred people in their 40s plus a few hundred all younger ages combined, some confirmed and some not, in a hard-hit nation of 67.5 million - is that really so hard to swallow? 

By-Country Analysis
Some other countries considered, just in text first, then visuals for some:
Spain: plotting shows excess for ages 15-44 at weeks 4-6, w12-14, solid excess at w5, 13 then average, then excess at w31 and 35, and all high near upper 2 in between.
Italy at upper 2 w8 9 12 14 16, and low since
France: w13, 15,  then a steep fall w 16/17, average since, until a rise to near upper 2 at w32
Austria: a smooth rise peaking w11 near upper-z, notable lack of low usual low weeks - all high must add up.
Belgium: rises within normal at weeks 5, 12, 36
Portugal: at upper 2 w27,29,31, with actual excess w29
Switzerland: a rise peaks at upper 2 w15, 23
Sweden scrapes it w10, 15, but otherwise near and often below baseline. This was part of their plan - it was the old Mr. Tegnell was willing to lose a bunch of.
Netherlands: odd pattern, possibly unrelated (more study in image): early spikes, moderate since, then near-excess again at weeks 33/34 and 36.

In my crude visual explorations below, I'll show the interrelated rises and falls between 3 age brackets (15-44, 45-64, and 65+), while exploring a new theme I noticed; the disputed death cause(s) seems to spread from the younger to older over time, with peaks tending to appear about a week later in each bracket up from 15-44 (with a week being the finest time unit available). In the first wave, each level also shows a higher peak of deaths and a longer post-peak high following it, and with little or no time delay. In subsequent waves over the summer and fall, the same effect appears but much more muted, sometimes invisible, and it takes longer for a spike of deaths to appear each level up. Those aged 45 and up and especially aged 65 up have by and large learned to avoid Covid19, and they do that now. They still die from it, but in lower number and after evading it longer than than they did before. 

England, noted and visual: w13-16 high excess, pretty high since, esp. w30, 32. w13-17 at upper 5, peaking at z-score 6.45 week 15 - solid excess past the 2 w20, slight w23, 30, 32, 38. Legendary excess among the older groups during the first wave. As often happens, this instill a strong communal lesson - from purple forward, the time delay between a surge hitting different age groups is longer than any others I looked at - 2 and 3 weeks, to the 1 or 2 weeks or no offset with the rest. If English youngsters are especially infected and dangerous, could it be the older folks know this and avoid them a bit better than usual?
There's also a notable rise in all ages deaths in the first weeks of the year, spiking shortly after Christmas family visits would occur, with ages 65+ in excess, scraping the upper 5 in weeks 1 and 2. A similar trend in the US involved a holiday surge of mysterious "pneumonia" deaths in week 1, just as the first confirmed covid19 death (unknown at the time) occurred on January 4. Did England see the same kind of thing?

Wales - the closest thing to another England, most likely to develop or catch the same problem: similar rise offset between ages but shorter - mostly one week, and note above how 15-44 deaths start below average/baseline prior to the April peak, and remains solidly above after that. Wales displays the same pattern, but more muted. As noted above, all EUROMOMO nations combined also shows the same basic pattern. From week 20 forward it rises to entirely above baseline, scraping the upper 2 at week 22 and surpassing it at week 33, while deaths for older classes remain close to or below baseline levels of death.

Spain: less time delay - generational mixing is the norm? (orange, green). But in the purple min-rise - is the loosened lockdown, continued segregation of elders? And that was later loosened? I haven't followed, but if I had to guess... 


Also I had noted before a surge of deaths in Spain in late 2019-2020, when they did have at least a couple of known covid cases. There on the left we can see a similar pattern between all age brackets. But just what caused that to level off and drop well before the bulk of covid cases would take some explaining. Maybe they had a bad flu before, that actually WAS its own thing.

Italy- looking milder than I'd think relative to the red line for upper-5 significant increase. Great display of magnified and delayed tolls at each bracket up, following a basic shared pattern in all three color-coded phases. 


France - as with Spain, not so much for the time delay, and the deaths among younger people tends to come at the end of each surge, when it's more often at the start.  But again rises in deaths occur in all 3 groups at a time of surging likely covid deaths, and to an increasingly similar scale (on the Z-scale) each time. See in France too how deadly lockdown was to 15-44 year-olds (weeks 16-19, most recent weeks)


Netherlands: almost backwards for age offset, but again all three ages brackets have surges of death around the same time, most likely from the same cause.

Blah Blah then End
People aged 15-44  - especially the younger half of the bracket - are on balance the most mobile and least cautious, the least burdened by a sense of personal danger, and thus the most likely to stumble into the SARS Cov-2 virus early on, and to be the main drivers of contagion everywhere. They don't die as often as older patients do, but they suffer some deaths along the way - especially among the older half, and primarily those aged 40-44, or anyone with the right conditions.  It won't then move to an older group an leave them alone, but the graphs almost look that way. Some of them may have some a heart valve disorder, etc. and would die suddenly, while less of them die after a struggle (absent some such swift assassination, they can often win a struggle), so their deaths might spike earlier in a surge.

Older folks know the danger and are more careful, but when it's spreading widely enough, many wind up getting it eventually -  around a week later on average, and much more fatally per capita. Many of these deaths are also swift, but it more often comes to a struggle, which they more often lose. It will be just as fatal to them now, but less are dying. This means one way or another, the virus is getting to less of them, and getting there later when it does.

The younger folks, by contrast, haven't learned they have anything to learn, and so they keep getting infected and thus dying at the same or even higher rates as they were before. Considering the lessons most of us have drawn, so many convinced this is a harmless flu except to people over 60 or 70 or at least over 50 - they WOULD get infected more widely, and find their fatal percentage much more effectively, apply it to a far greater number, and gradually or swiftly skew the average death age younger. 

If we could only see just where in 15-44 it's centered, we could probably see even more clearly how my reading makes the most sense - ignorance is the #1 extra contributing factor to this issue. 

England is heavy on semi-misinformed libertarian lockdown rejectionists and herd-immunity-enthusiasts, even among otherwise bright people. It does not surprise me that an unusual proportion of the 15-44 drivers there wind up dead, with both higher spikes and a higher prolonged elevation between. I'm not surprised to see it bleeding over into Wales. All things considered, this probably is due to the same virus killing so many others. More likely there are just far more people aged 15-44 than we realize, far more than normal, with unconfirmed infections. Many will have sought out infection deliberately. The unconfirmed fatality ratio we'd be looking at might then be about the usual. Maybe it higher than even this, because of more diabetes than usual, or whatever. But either way, it all makes sense to me. 

Here's another thing I found to help understand: ONS 2019-2020 weekly all-cause mortality surveillance for England and the "devolved administrations" of Wales, Northern Ireland, and Scotland, in less detail and the latter with a 2-week time lag.  


upper 2 (upper normal range) used here, and broader age groups: 0-15, 15-64, and 65+. Just one group sees excess past week 23, and it includes the segment in question, adding a bit of time-and-space detail to the picture. 

We'll start at the last high week you could call part of the April peak - week 22, ending May 29 (example PDF link): "In week 22 2020 in England, statistically significant excess mortality by
week of death above the upper 2 z-score threshold was seen overall. Statistically significant excess mortality was seen by age group in the 15-64 and 65+ year-olds and sub nationally (all ages) in the North East, North West, Yorkshire & Humber, East & West Midlands, East of England, South East and South West regions." No excess deaths in Wales, Northern Ireland, but  (as usual noted in a later report) there was in Scotland back in week 20, when almost everywhere was.  

From there forward, statistically significant excess mortality was seen in ...
week 23: Excess seen sub nationally (all ages) in Yorkshire & Humber and East Midlands regions (not Scotland or anywhere outside England, from here on out). 
week 24: no excess anywhere
week 25: only in East Midlands (all ages)
week 26: East Midlands  (all ages)
week 27: no excess
week 28: ages 15-64, East Midlands 
week 29: ages 15-64, East Midlands 
week 31 no excess
week 32: ages 15-64, East Midlands  (all ages)
week 33: ages 15-64, East Midlands and South West  (all ages)
34: no excess 
35: no excess - 
36: "Statistically significant excess mortality was seen by age group in the 15-64 year-olds and sub nationally (all ages) in the North East, East Midlands and South West" - bit the actual death toll given is far lower than the other weeks - see notes below.
37: no upper-z excess 
38: no upper-z excess - ??
39: sub-nationally (all ages) in the North East and East Midlands regions

"All ages" doesn't mean every age group was in excess. It means ... not sure. Ages aren't given for reginal overage? General excess, but only a tiny bit over in a few areas, under in others, with no one bracket standing out for excess? 

Anyway, no age bracket stands out then except 15-64, apparently meaning nationwide, but clearly driven most by constant excess in East Midlands. It keeps doing that. Why? I offered my guess already, a couple of times. Sorry.  

So England's excess deaths also centered in the East Midlands, besides being younger than usual. Nearly all of them will have some kind of condition, even a mild and totally unknown one, been smokers, or obese. I bet they also have a higher tendency for taking their health advice from alternative media. Maybe just the person who infected them did. For whatever reason, the excess deaths never stopped there, hovering above and just below that generously high upper 2 z-score level. weeks 28-33 and 36 is when more younger folks start dying even more, which would be some days to week(s) after increased infections during the second wave in September. 

Ref: week 22 where I started, with deaths coming out 757 above my calculated multi-year average for that week. That's down from  +1,578 in week 21, following a peak of 12,711 above average (well over double the usual deaths) in week 16 (and week 17 wasn't much better). Those all labeled as showing significant excess deaths in all or most areas and age groups. But at week 35 where it's 857 above my average, it's called not significant. Then the next week it says deaths are up all over, despite the lowest numbers yet? Week 36: "Statistically significant excess mortality was seen by age group in the 15-64 year-olds and sub nationally (all ages) in the North East, East Midlands and South West" coming to the furthest below average toll yet: 1,807 below my 3-year-average. (7,739  deaths vs. 3ya 9,546 - ONS uses 9,182 5ya for this week). The pattern plotting these weeks 35-37 is \/, while the description is opposite, and EUROMOMO data for England agrees, giving the weeks a  /\ pattern, but kinf of tilted or flipped (and actual numbers unclear for better comparison). I'm still working on a big project involving all this that got hung up on that very confusion.

Update October 7:  
Figure 3 shows "Non-COVID-19 deaths are generally below the five-year average for younger age groups from Week 12 (week ending 20 March 2020), particularly those aged 10 to 39 years." 
notice the cut-off age. 40+ aren't so particular in being low, and they would grow in time. But they show weekly deaths by 10-year brackets, and on review, I may have overrated the 40s. At least by variation from the norm, the 30-39 bracket did almost as poorly, though compared to probably lower norms.
It seems plenty in their 30s die as well, and that's part of why 15-44 year-olds die so much.
Figure 2 in the later report shows the same over lockdown and into the summer, used here (remixed with notes) to compare the 4 age brackets where covid-age deaths can be still be seen as the ones in excess, a bit younger than you'd think, if not younger than usual.



Sunday, September 27, 2020

Tracking the UK "Second Wave" Death Toll

Adam Larson (aka Caustic Logic)

September 27, 2020


As the United Kingdom sees a rise in Covid19 infections, PM Johnson and others have suggested a new round of stay-at-home "lockdown" order may become necessary. Many are frustrated and confused. For example, independent journalist Anna Brees, a lockdown skeptic, shows the UK deaths curve (still looking totally flat at the end, and asks "lock me down for this?" 

As I see it, the first round of lockdown was "for this" - to make the people make that the death curve, instead of the one they were on the way to. As this shows and everyone knows, daily deaths had been doing nothing but rising sharply before they leveled off and then fell rather suddenly, starting in mid-April. That's about 2 weeks after restrictions were imposed and new infections witnessed a similar, if milder, decline. Most logically that's less contact, making less infections, and thus less fatalities. If that had been delayed, deaths would have kept rising to even higher levels it would take longer to arrest and then climb down from. 

Of course, it's a "second wave" and the deaths from that they worry about, and these haven't happened yet, aside from perhaps the low-rumbling start of them. I had just been looking gain at UK deaths again and noticed they've been rising - they had been down 4 or 5 most days, sometimes zero. But now they've gone back up to 10, 15, 26 a day and rising, with revisions pushing tallies into the 30s, and the latest full day currently says 45 deaths just in England and Wales on September 23 (these are Public Health / NHS stats - ONS still tallies a higher number for England and Wales, I noticed - 51,917 there, to official UK total 42,060 including Scotland and Northern Ireland). 

That rise doesn't yet show up clearly on a scale distorted by highs around 1,000 deaths per day, but it's a bad sign, and just the start of what will wind up hurting. Having a look at recent infections, it's clear that second wave is real and has already been happening. Official stats when I checked a few days ago - note the last day is tiny, provisional - all these days at the end can change, mostly up and probably just by a bit. With that, it's been hitting up to 5,000 confirmed cases per day. (in fact one day now revised to 5,197, and the site currently says for daily cases not yet plotted: 6,042. They will need to expand their graph.

https://coronavirus.data.gov.uk/cases


Maybe all those freewheeling protests are having an impact? Maybe some herd-immunity vigilantes are even out there deliberately spreading it to save the UK from tyranny? Possible. But mainly it's just the communicability of this virus plus a lot more communing than was happening; regular commerce and social activity. But then see the photo below from Trafalgar Square yesterday, September 26 (Time). To the extent this spirit prevails, their activity is not being done carefully except where it's required - and it largely has been, which is what's got these people so frustrated they prove as loud and clear as possible that they need to be babysat, even as they demand not to be. Babies, after all, don't understand.



Already confirmed infections are near past April levels. That's partly due to wider testing, of course, so it's not really that big in comparison - yet. And I do not foresee near-April levels of deaths coming on (above 800/day for a month straight back then), nor even the same rate scaled down to current real infections. "Protecting the vulnerable" can never be done totally, but it has been done fairly well, and that's been the main thing, IMO, that lets deaths remain so low people can easily pretend the danger has gone away for good. But again, it can't be done totally, and with this scale of infections that we know of, more yet that we don't, and especially if higher levels are yet to come -it will be harder for those of us in serious danger to keep avoiding this extremely contagious pathogen.

If second waves happen for real and not just in British leaders' agendas - maybe France would see one?  2nd wave - cases now far above April levels, confirming up to 16,000 cases in a day, and deaths are back up to 40-50 a day since Sept. 15, already with three spike-catchup days of 80, 154, and 150 each (in a slightly smaller populace than the UK). https://www.bing.com/covid/local/france?vert=graph


Now the UK death toll remains almost flat so far, but of course it takes time to die from the virus, so trying to call the second wave death toll is clearly premature. Almost instantly on seeing cases near 5,000/day, I predict the daily deaths will rise to 50, then 100, then hitting triple-digit daily tolls regularly, just from what we see here. Depending how high it gets, they may again see 300 dead in a day or more. Depending how long it takes to rein that back in, it could add a hell of a lot more deaths along the way. 

Projecting 2 Weeks' Deaths

Premature or not, I tried for a data-based estimate, and was able to show my initial guess  (" triple-digit daily tolls regularly, just from what we see here") was high. I tried to line up dates between cases, hospitalizations, and deaths in various ways - something I hadn't tried yet and didn't know how to do. For reference, I used this drugs.com page with a timeline of general progression that lines up with the bits I've heard: first symptoms is day 1 - fever, cough, aches, etc. Around days 7-9 dyspnoea and sepsis set in, in bad cases. Day 12 or so, improvement begins or ARDS forms, patient is admitted to ICU.  "Day 15: Acute kidney and cardiac injury becomes evident" and then "Day 18.5: The median time it takes from the first symptoms of COVID-19 to death is 18.5 days." (which would make it day 19.5, I think, starting from 1)

First I rushed off with a comparison of cases to deaths 18 days later (giving a crude cfr of 1.28% for what it's worth). But confirmation doesn't happen at infection. I suppose just when varies ... but the above gives 6 days on average in ICU, in turn usually 12 days after infection.  ... I tried a lineup of the curves for cases, hosp, deaths, on the idea surges would tend to show up similarly at the right times, besides scattering to surrounding days and mixing with other days' scatter. But with all the local wiggles and uncertain variables, it was hard to say. Generally, having the three set about  a week apart makes for a basic fit. Somehow I decided to use: hospital 6 days after confirmation, death 9 days after that (just 15 days total span). A ways in I realized even this wasn't the best, but I don't have the combination of skills and time to get it right, and it doesn't matter much - the differences are only so big, and some will cancel each other out.

Over the most recent 23 reasonably complete days, I count: 

372 deaths up to 9/23 (from 9/1), out of 

37,979 cases up to 9/8 (corresponding days), and 

3,207 hospitalized up to 9/14 (corresponding days).

I calculate a 0.98% crude CFR (case-fatality ratio) - nearly 1% of known cases have died so far. That's rather low and more like an Infection-Fatality ratio (IFR), historically, but confirmation is probably much higher now than it has been, bringing IFR and CFR a bit closer to synonymous.)

8.44% hospitalization rate - some 92% are considered ok to quarantine at home, but 8% require intensive help

11.6% crude "hfr" (hospitalization-fatality rate) between the 23 paired days (deaths 6 days later) - nearly 12% who have to go to the hospital never leave.

Using this cfr and hfr compared to daily cases and hospitalizations from Sept. 25, I project a total of around 519 deaths set for the next 14 days, as broken down below. 

The "cfr" I get comparing deaths 15 days from cases seems fairly steady down the line, but my hospital fatality rates are increasing over time. Around Sept. 5/6 hospitalizations is where it starts rising faster, from about 9% (as set up) to as high as 17.7 and 19.5%. An average of last 12 days gives a 13% hfr, which seemed to better fit the actual rise in deaths. so I used that. In fact the latest death tolls might show I'm still estimating low. 

Hospitalizations themselves seem to increase in the same time, from 130/day for a week up to Sept. 7, then quickly to 164, 192, 192, 205, 198, 208, 231, 247, 248.  Why these rise? wrong day line-up causes an increased distortion, or was causing it, and the higher rate is more accurate? Or hospitalizations and the fatality rate of them are actually rising? It could be that as young and careless people people spread the virus wider, it's now be spreading more to the less eager citizens; they avoid it for so many days and then don't, wind up in the hospital at a higher rate, and then die there at a higher rate than we were seeing 2 weeks ago. 

With this set-up, the longest projection is 15 days out from the latest reported cases tally (again, subject to change). I used 0.98% for each day's cases, and 13% for each corresponding day's hospitalizations (where available - six days later runs out 6 days before the end), to estimated deaths 15 days out. I just averaged the cfr and hfr predictions, which tended to be quite close. The 13% hfr does pull it a bit higher than the cfr alone, adding about 4-5 deaths at the end. So I mimicked that, adding 4-5 from there. We'll see if the current hfr makes it lower or higher than I estimate. This suggests what my earlier try did for a peak of deaths so far at 55 in a day - and I emphasize - so far.


What I'll be Watching

I will come back to check my prediction and bring updates on the confirmed death toll, the reactions of the government and the public, and maybe a look at the unconfirmed death toll. Another post deals with overall UK "excess deaths" (besides sorting out the varied and changing official tallies) - speculative analysis of the true death toll based on when and how people die at above-average rates. many have noted some 30,000 disputed non-covid "excess deaths" above the average, disputed if caused by the virus, or lockdown, or what. It tends to happen right when Covid19 is killing a lot, and when we can see how, it's in mostly cardiovascular ways where the virus might well play a hand. I'm having another go at this, expanded in a few ways, including to track the recent rises in covid and overall deaths, now above average when they had been well below for about 2 months. And I'll bring some more specific clues that bear out my thesis those excess deaths were probably more coronavirus ones never confirmed - and likely even some of the deaths BELOW the average line are as well. I've found so far the true U.K. death toll is probably over 70,000, possibly  as high as 80,000. I'll see about refining that point at that post,  after another week might settle some trend confusion I'll have to explain - maybe after I understand it.  

October 19: I haven't followed all developments. But cases are now at around 18-20,000 per day, hospitalizations are 900-1,000/day, and deaths are higher than I predicted, hitting around 90 most days and so far maxing out on October 11 at 103 (England and Wales - 117 for all U.K.). As before, all days open to revision, mainly up, and especially in the last few days.



Saturday, September 19, 2020

United States: Covid-19 and 4,000+ Extra "Influenza/Pneumonia" Deaths

September 19, 2020

I've been doing some excess and by-cause deaths analysis that's been especially fruitful regarding 2020's weekly fatalities in the United States. So far this suggests - as many others have noted - many thousands of deaths probably caused by unconfirmed and unknown infections by Covid19/SARS 2. I did some tracking of mysteriously surging deaths from heart disease, Alzheimer's, diabetes, and more, using a CDC data table, updated weekly to show weekly totals for many natural causes, by state and national, for both  2019 and 2020 for comparison. That is available here:

https://data.cdc.gov/NCHS/Weekly-Counts-of-Deaths-by-State-and-Select-Causes/muzy-jte6/data

The latest weeks are always incomplete, and weekly data continues to be revised to varying degrees for many weeks afterwards. So this is provisional data, but so far it seems fairly complete and settled over most of the long span examined.

Those other categories might seem unclear in their relation, but the SARS2 coronavirus causes circulatory and other damage in many of those it infects. That will include ones we didn't know were infected, who for example had heart disease to start with, then die soon after an unknown illness maybe even they didn't feel (more likely they just didn't say anything). Case-by-case there might seem no reason to guess covid involvement, but when you see the scale and timing of the rises, it's hard not to see. (a rough sketch of several of these is included at the bottom of this article).

The category I'm looking at now is a little more direct in its relation: "influenza/pneumonia (J09-J18)" - for short I/P. I may be hazy what's included. The related codes are broken down elsewhere as:

J09  Influenza due to certain identified influenza viruses
J10  Influenza due to other identified influenza virus
J11  Influenza due to unidentified influenza virus
J12  Viral pneumonia, not elsewhere classified
J13  Pneumonia due to Streptococcus pneumoniae
J14  Pneumonia due to Hemophilus influenzae
J15  Bacterial pneumonia, not elsewhere classified
J16  Pneumonia due to other infectious organisms, not elsewhere classified
J17  Pneumonia in diseases classified elsewhere
J18  Pneumonia, unspecified organism

Then I'm hazy on the exact definition of influenza vs. pneumonia. But anyway ... I think it includes the two main classes of influenza (A+B) and a larger number of deaths caused by pneumonia, which I gather has many causes (not a specific virus or bacteria but many of both)but - as I gather - not the fullest list of pneumonia deaths (some may be in other categories here). I suspect Covid19 or "Kung Flu" as a select few have called it, would fit here, as a pneumonia or maybe a flu of unknown cause, until it was given its own category on account of being this deadly super-bug. 

And when people don't know it's covid19 they're seeing, it may keep getting listed here.

I had seen some evidence cold and flu transmission dropped sharply under lockdown conditions, in the U.K. anyway. That makes sense, and I expected lower cases and thus deaths in most or all places that used lockdown and advised social distancing, etc. That includes the U.S. of course, even though mitigation efforts were not universally employed and complied with. In contrast, 2020's I/P death toll is high and erratic, in some states I've looked at (but not others), and on a national level; U.S. totals include quite a few hundred more than expected. 

On analysis, these deaths seems to be below average for a baseline (or below 2019 levels anyway), and by when the extra deaths come in, it's almost certain they were caused by unknown Covid19 infections. Perhaps the most interesting detail is how the biggest spike of these comes right at the year's start, immediately after Christmas and New Year's, suggesting the virus was already circulating widely by then, about a month earlier than known.

Here is the graphic, with some discussion following. It's too small to read in this preview, except that:
* the tall red curve is Covid deaths (mostly far off this scale to 2,000), 
* the darker orange line is 2020 I/P deaths, 
* the lighter orange line running normally beneath it is the same thing in 2019, and 
* the steep green slopes show the seasonal declines in flu cases for for each year, with 2020 being the darker, steeper, and earlier of the two (explained below).


To get a clear picture, I had to trace the data back to late 2019 to be fairly sure I was seeing a pre-covid baseline. For comparison purposes, I tried a bit to find the equivalent I/P tallies from late 2018. One list broke down flu (tiny numbers like 30) and pneumonia, with bigger numbers in the 3,000 per week range - far bigger than the kind of combined totals I've been plotting. The same number included covid-19 deaths in the pneumonia category when they came in, but it started out higher even before those. It must include other illnesses lumped differently than on the table I'm using, which may have their own patterns, but ... for now I'll stick with the 2019-2020 table and the one column, and offer a guess for the late-2018 lead-up, traced back from the visible trend, with a reasonable seasonal increase. (this lack of comparable historical data is also why I didn't even try for a multi-year average - I just use 2019's weekly levels, using a relatively normal year as a stand-in for an average.)

2019 ends with around 8-900 I/P deaths per week, well under what late 2018 probably should be, climbing past 1,000 in a rise that might be all seasonal, or include a few early covid deaths. Then 2020 starts with a jump to 1,555 deaths in week 1, ending Jan. 4. That's 283 more I/P deaths than in the same week of 2019, when it had been about 100-200 per week below that level.  

This is unusual, and it comes just as the first few confirmed Covid19 deaths were coming in. As listed on the CDC table, three deaths occurred prior to the first confirmed infection we heard about at the time, on the 21st. The first listed covid death is on January 4, at the end of that deadliest I/P week. (2 more came the following week, then alternating 1 and 0 for a few weeks before a quick takeoff in weeks 10-13.) 

As of mid-December we were just hearing about a deadly flu in China, but it seemed like a Chinese issue we had some distance on, and it was somehow not certain if it was even contagious. Later we learn it was contagious and was a Chinese problem from October or earlier, and could have spread anywhere since then. And seemingly out of nowhere, some extra-deadly infections occur somewhere in the United States roughly at the Christmas and New Year's holiday. That's a time known for visits to older family members, a few spare days for incubation, and people packed on airplanes flying to airports across the nation and overseas. Note that very frail people can die from the virus and the damage it causes within just a few days rather than the common struggle of 2-3 weeks. It can be incubation then almost instant death. Infected with SARS at Christmas and died New Years to be classed as an I/P death ... may have happened about in about 280 cases nationwide last year. That would be almost a one-in-a-million occurrence, when it had seemed like something that couldn't happen at all.

The elevated I/P death toll stabilizes after the January jolt, but remains well above 2019 levels to week 7, and roughly matches 2019 in week 8 (ending Feb.22). Then deaths rise steadily, just as weekly Covid deaths shoot straight from 9 to 35, 53, 571, and 3,155 over weeks 9-13. I/P deaths peak just before Covid ones, with 1,870 logged in week 14 - 646 more deaths than in week 14, 2019 - in fact an increase of over 50%. 

An interesting detail is how test-confirmed infections of influenza fall off here, a month earlier and more sharply than in 2019. See below bar graphs of weekly test results for the last 4 seasons, broken down by influenza types. The latest is from the weekly CDC update - but that changes, and where it'll be then, not sure. The others from prior years are from quick Bing image searches. 

For my graphic, I took the two relevant graphs (2018/19 and 2019/20) into my graphic, roughly scaled them vertically to my scale, drew a shape (in MS Paint) including the green line between bar tops, then set and stretched it between the right weeks on my timeline (and erased the rest of the shape - I like the tedium up to a point).

So 2020 has a rather sharp decline in flu cases just as Covid19 was taking off. However, by these graphs, infections always fall pretty sharply at around weeks 10-20, and in 2020 testing/confirmation may be low due to "lockdown" and such, which is when the numbers drop so sharply. But also transmissions should be down as people kept a distance from each other. It might be the vast number of actual flu infections dropped just as sharply as these few that ever get confirmed. 

The same should probably apply to all kinds of pneumonia - if it's contagious, it should suddenly spread more slowly under lockdown and social distancing, with the use of masks, etc. Yet something classed as "influenza/pneumonia" - that might be especially deadly and/or extra-contagious - infects enough people to push the death toll far higher than usual. When it strikes just as Covid19 was taking off, there's only one overwhelmingly logical culprit.

Back to my graphic and its narrative: supposed influenza/pneumonia deaths peak in week 14 (ending April 4), then decline sharply by the following week. The infections leading to those deaths must have fallen off several days to a week or more before this. This would be roughly in late-March to early April, which is when most states imposed strong control measures, and president Trump did, to his credit, send out that postcard to everyone, including his hardcore fans who had been doubting the danger, based on some of the conflicting things Trump had said. From here the weekly deaths decline steadily to roughly 2019 levels (around 800/week declining to around 650) by week 19. This still probably includes some Covid deaths and thus less of all the other kinds than in 2019. 

A week after the 4th of July, the flu and pneumonia get a bit deadlier again, just as Covid does the same, but much milder than after the surprise attack of January. The level only rises to slightly above average, and holds about there for 3 weeks before falling again to seasonal levels and even below 2019 at several points. But it should be a bit higher at the end than shown here; final data points tend to be incomplete (the last two so incomplete I didn't even include them). 

And again, this is after after I/P deaths started - and likely stayed - below 2019 levels, aside from the stray coronavirus deaths that probably explain all this erratic overage. Just through week 18, there were 3,988 deaths above 2019 levels. From there to week 34, the ups and downs combine to just 51 deaths above. 4,039 total. But it seems 2020 was a low year, with a baseline of something like 100-200 I/P deaths per week below 2019 levels. That would add about 1-200 extra, likely-covid, deaths per week to explain the levels we see. As the increase only starts clearly in week 1, that math is easy - it would be between 3,400 and 6,800 extra deaths added to all the differences from 2019 levels. including any baseline adjustment for now. I'm pretty sure there is one, but the late-2018 level being unclear keeps me from  being able to say the scale of it, except that I can safely say it will add at least another 1,000 deaths. Conservative total: More than 5,000. 

This adds to the notable but yet-to-be-calculated overages in deaths from (at least) heart disease, Alzheimer's, diabetes, and unclassified causes (R00-R99), which also can relate to unknown covid19 infections, and swell just before, after, and during spikes of covid deaths. Just heart disease is over by more than 2,000 deaths in the worst point, weeks 15, just before covid deaths also peaked at 16/17,000. Alzheimer's and diabetes were both up about 600 deaths the same week. To be revised, but here's my prior US total analysis, with numbers that have been changed since, and including a rough version of this I/P pattern, from week 4 on (where they differ, the new one is both updated and more carefully done). Some of these other death causes had showed high early levels I mostly ignored, but after re-examining the New Year's I/P deaths, I should extend this back further when I revise (after the next weekly updates, if they come).



Thursday, September 10, 2020

On the Covid-19 1% IFR "Blunder" that's Actually the Truth

Adam Larson (aka Caustic Logic)

September 10, 2020

additions on: 9/11, 12, 14, 16 ...


https://www.rt.com/op-ed/500000-covid19-math-mistake-panic/

I call out RT and the author to Malcolm Kendrick ("skepitc, doctor, writer, blogger" - on Twitter) to retract this nonsense. We've had more than enough of this already, but the usual folks are loving this -yet another supposed proof almost no one died from Covid-19 now that it's over, nor was much of anyone ever going to die, except of course from control measures. As his nodding fans all know, this "lockdown" stuff was a horrible mistake, if not an evil plot.

Kendrick argues here a simple math mistake no one double-checked was the basis for the global lockdown reaction - which was already pioneered in its most extreme form by China before the alleged mistake. But anyway ...  It's a short enough read I won't summarize it first, just in explaining how terribly wrong it is, below. If you aren't clear on CFR vs. IFR, it explains that.  

Kendrick: "Covid’s CFR was likely to be about one percent, so one person dying from a hundred who fell seriously ill. Which, as time has passed, has proved to be pretty accurate." False. The global average Case-Fatality Ratio (basically % of KNOWN infections who've died) has usually been around 5% - US is around 3% - almost everyone is over 2% and some are past 10%. The UK currently boasts 11.7% (as currently listed at the Bing tracker). This UK doctor thinks it's less than 10% what it actually is. Because skepticism?

"Almost no-one is dying of Covid anymore" he says. Because of actions including a slowed spread, deaths in the UK have gone from terrible to mild, nearing zero some days. Like many such skewed thinkers in the UK, he conflates this (along with a mythical reading of Sweden's experience) into a global reality. Global daily deaths, with spikes and dips from reporting variance) show the current numbers are almost on par with the worst days. It's been far better in those places hit early, but the virus has been getting its foot in more and more doors worldwide since then. 



So the virus proved harmless after a bit for some reason, as Kendrick concludes, and he then goes on to complain:

"But because we panicked, we’ve added hugely to the toll. Excess mortality between March and May was around 70,000, not the 40,000 who died of/with Covid. Which means 30,000 may have died directly as a result of the actions we took."

I have mapped this out, as shown. Kendrick could not say how these thousands of deaths happened so quickly atop the thousands of covid ones, then mellowed the same way. In the US, we can see the excesses are mostly in classes of death consistent with unknown Covid-19 infection - cardiovascular in nature, and spiking just when Covid deaths do, just before and after, etc (see here). I say this is 70-80,000 UK people dying almost entirely from Covid-19 infection. 


Excess deaths all over suggested the mark of one million dead was passed some time ago. Now it's officially past 900,000 and I estimate 1.6 to  1.8 million actually killed so far. Annual flu deaths range from around 350,000 in a light year to 650,000 in a bad year, with little to no intervention slowing it. This has killed at least 150% of a bad flu year, almost entirely in the last 6 months, despite massive and bemoaned efforts to slow it. Kendrick ... claims to still believe the early predictions it would kill no worse than a yearly flu. He seems to believe it has killed, and will kill that way. He's not alone here in being grossly out of step with reality.

Kendrick says (thinking UK and expanding it to the world) "no deaths were prevented." None. And he's thinking UK wrongly. He thinks the virus stopped killing at those exponentially increasing rates until about 2 weeks after lockdown was imposed, then fall sharply to a low murmur you can think of as gone ... all on its own, coincidentally if not mockingly right when people were trying to make that happen. 

As always for such fantasists, he insist on using past-tense phrasing: "no deaths WERE prevented," not "have been." Covid "was never" going to kill many, etc. He's probably been one of those calling the crisis over from the minute it started. The man goes on to make a number of other dubious claims I'll skip to get to the central point - the predictions of a 0.1% IFR (infection fatality rate) that Kendrick says have been borne out, vs the deadly, lockdown-justifying tenfold exaggeration, which was "horribly, catastrophically, running-into-Mars-at-5,000-miles-an-hour wrong." 

However ... If the 0.1% estimate was actually blundered to 1% (I didn't re-check his reasoning there), it's sort of interesting, and a rather lucky turn, as a 1% IFR roughly predicted the reality to come. It seems the early guesses were the blunders here, and ones with deadly potential if people were to believe in and act on them in the current reality. People in the U.K. especially are hungry to be so misled, and this smug doctor Kendrick has fed that, just casually splashing more fuel on that fire of deadly ignorance. 

Oh, and he's a doctor, so he must know, right?

Let's start with his sources. It seems that 0.1% estimate came from an early statement by the CDC's Dr. Anthony Fauci in February - one he clearly doesn't stand by today. Kendrick buys and leases it out anyway, even boosting it to a maximum estimate: "Covid was never going to kill more than about 0.1 percent – max." He also cites another February prediction from the CDC that mused the virus' effects would be "akin to those of a severe seasonal influenza,” where only something like one in a thousand will die. The scandal, to Kendrick, is how those same experts soon came out saying it would be ten times worse - 1% were likely to die. They didn't stand by the first guess for long at all. Or maybe they did just conflate CFR and IFR and bungled what they meant to say - maybe on purpose. I mean there clearly is some plan to all this exaggeration and fear-control, right? They wreck the economy, get us in masks, and viola, every dictator's wet dream - people in masks and a wrecked economy. (it's a complicated plot, I gather)

This is supposed to be our glimpse of the truth? Two seemingly retracted early guesses and a supposed mix-up? 

Kendrick says the facts bear it out. But if just 0.1% of infections - or anywhere close - result in fatality, the following would not already be established facts:

* Bergamo, Italy: city and province have around 0.55% of their POPULACE already killed. That means even if everyone had been infected, it PROVED about 5-6 time deadlier than Kendrick would have you believe. Most of the 6,000 killed were by early June, when serology tests showed about 56% had antibodies. (see here) That lapses but not much, so let's say 58% infection rate.

So Bergamo estimate IFR = 0.95% - not 0.1%. 

* CDC report: In Orleans and Jefferson Parishes, Louisiana, "The overall IFR was 1.63%" (Reported deaths were divided by number of persons presumed to be recovered to calculate the IFR) 

- "IFRs found in other seroprevalence studies are lower, ranging from 0.5%–1.2%"

- No mention of 0.1% IFRs here. Some bunch of blunders? More CDC stuff you can only trust up until they started getting more facts in March? 

* New York City Covid-19 deaths: 23,741 (conf. + prob.) - est. by excess deaths: ~25,500 (quick calc.). So just among the city POPULACE (8.4m), not the infected ... something like 0.28% to 0.3% of those once alive in New York City HAVE - ALRAEDY - DIED from the coronavirus - not 0.1% "max" of those infected. 

Similar death rates have been recorded in places like Madrid and Rio De Janeiro, and Mexico City looks especially bad by excess deaths. Those three cities are a mixed-bag for lockdown measures, but all have used at least some control measures people like Kendrick would complain were destructive.  My data is behind or fragmentary, but last I checked  all three come out somewhere well past 0.25% of the POPULACE already killed - not 0.1% of the infected. 

Guayas province, Ecuador, est. from excess deaths of 13-14,000 (as reported just by the start of summer) say close to 0.35% of the POPULACE has died. That will be higher among the infected, and both rates will be higher in their temporary capitol Guayaquil (it's a mess and data is scarce). 

Back to NYC: Among those infected ... est. infection rate 28-30% (just over 27% had antibodies in August - NYC study - NYT report) = 2,352,000 to 2,520,000 estimated infections. Therefore NYC est. IFR = 0.94% to 1.08%

That's eerily close to what we see in Bergamo AND to what that "blunder" had said. Kendrick moans how this error - that turns out quite correct - is what prompted nearly every national and local government on Earth to respond in these drastic ways - after consulting their best expert minds, and just as many of them were dealing with raging infections, swamped hospitals with sickened staff, and unchecked surges of fatalities presaging far more than 0.1% were in serious danger.

Then someone made a typo, and they all decided to lock down. Huh. From Shanghai to Tehran and beyond, including before the error was made. Huh. 

So ignorant predictions of February suggested it would be flu-like, but the facts since have clarified the IFR is frequently as high as and higher than the "blunder" estimate - IN FACT about 1% of those who get this virus so far have died. That average includes the deadly early attacks and should fall in time, unless people gets careless enough, or the virus mutates. But it'll stay far closer to 1% THAN TO 0.1% "max." I'll let Kendrick put the 1% reality back in context - he did that well when he was calling it a fatal error:

"You may not think that percentage is enormous, but one percent of the population of the world is 70 million people – and that’s a lot. It would mean 3.2 million Americans dead, and 670,000 Britons." 

Yeah, if that were happening, that'd be terrible. You'd want to know and take it seriously, right? Because something like that ... even if you slowed it down with drastic measures, but then people resisted them left and right, you might have about 70-80,000 dead in the UK by 6 months in (about now). You might have 250-300k in the US, 900k confirmed worldwide and likely 1.6-1.8 million dead six months in, just thankful it hasn't been worse so far. You might be annoyed with idiots who dismiss the danger with this kind of lazy, self-centered "skepticism" that threatens us all. 


9/11 add: Will Merthon on Twitter with a detailed argument even the 0.1% mix-up never happened - Kendrick  passed on a flawed bit of deduction by Ronald Brown published by Cambridge University, involving his own mix-up regarding influenza mortality. P. Hitchens asked if anyone dared respond to Kendrick's "terrific article." I suppose he'll ignore my response.

9/12: People are pointing to Iceland as the low-IFR standard for the world - not Bergamo, Madrid, Louisiana, New York City ... It's said they have (or had) one of he best views of infection rates via widespread testing. If so, it adds some weight to there being only so much higher than 0.63% infected (2,161 confirmed cases to date, pop. 341,243). It might be five times that, but just comparing known deaths (10) to known cases, basic CFR: 0.46%. That's low. Few nations have anything below 1-2%, and most center somewhere near 5%, ranging not much past 10% in worst cases.

The best way to find the actual death rate - in Iceland as anywhere - will not be to compare confirmed AND unconfirmed cases with those same confirmed deaths (the basic CFR-IFR correction used at the Kendrick level). It would be fairer to compare confirmed AND unconfirmed deaths to the same totality of infections. Right? Apples AND oranges compares to apples AND oranges. This is excess deaths type analysis. I'm having a look into this and will report back. 

I noticed the Louisiana study via Dr. Eric Feigl-Ding Sep 6 - he's got some combo of knowing his stuff and having some kind of anti-virus bias for some reason. He also directs us to a re-analysis of Diamond Princess cruise ship outbreak - a case I might have recalled above. A confined and older-skewed population yields a higher-than-usual IFR of nearly 2%, or - mainly - they suggest past estimates were low, based on flawed methodology and then being too decisive or getting taken too literally. In contrast the authors here are methodical and I didn't get to just what they conclude.  2.0% does come up. previously I had: 3,711 passengers and crew - 712 cases - 13 deaths by most reports, some say 14 = 19% infected, and an estimated IFR of 1.83% to 1.96% on the Diamond Princess - even worse than those areas in Lousiana, about 2% of those infected on that cruise wound up dead. NOT 0.1%. tops. Not the 0.3% tops this herd immunity enthusiast adduces from some reading off of Iceland. 

About 1.9% matches the unusually clear (if not definitive) number of cases (and deaths!), and Fig. 1 in the analysis notes "combining the age-stratified nCFR from China with the age profile of Diamond Princess patients, 15.15 deaths should be observed on the cruise." How that last 15% of a death would be observed in unclear, but  that's about what I got in another way. IFR in China ... that's something else to check on.


Add 9/14: More on the IFR that has no one answer: It can get this bad:

* IFR = 30.86% at Life Center in Kirkland, Washingtomn, USA - of 120 residents, 81 (67.5%) were infected and 25 died.

* IFR among people 80+ in Lombardy, prior to mid-March control measures: 30% https://swprs.org/studies-on-covid-19-lethality/

* Belgium nursing home IFR est. 28% to 45% (ibid)

* 2% Diamond Princess

* 1.63% in 2 parishes in Louisiana (1.72% for African-Americans there) 

* 1.4% in Conn. to May 3 (Reason)

* Spain: 0.82% (based on confirmed Covid-19 deaths) and 1.07% (based on excess all-cause deaths).

https://www.medrxiv.org/content/10.1101/2020.06.20.20136234v1

Note some of the are from a useful article at Swiss Policy Research that considers in detail places with high levels like 1%, agrees with it, but suggests there may be 2-5 times as many cases as seroprevalence studies suggest. NYC may be 50% infected or more - Bergamo 114% or more? Again and again he notes "people with anosmia (temporary loss of the sense of smell or taste), a very typical Covid symptom" tended to not have antibodies, suggesting others without antibodies also had it, and maybe a huge portion that would drive down the CFR (twice as many cases to the same number of deaths halves the death rate). For the moment that seems possible, but I've also read than antibodies hold in a vast majority of cases, not in 1/5 to 1/2 of them. 

As for Dr. Fauci's initial 0.1% guess - as WebMD reports: "Fauci and other public health experts have since put the COVID-19 death rate at about 0.6% -- six times that of a typical flu season -- which is the latest CDC projection. “It looks like now that the fatality rate of a person who gets infected with this, on average, is around six times that of the seasonal flu -- so around the 0.6 [percent] range,” say Amesh A. Adalja, MD."

That's lower than what I see - in especially deadly places - but not by a whole lot. And others are faring better, bringing the average down.

* IFR CDC cited range (0.5%–1.2%) refs not openly available, refer to studies in Spain, Switzerland, Brazil. And I suppose it doesn't have to bottom out at 0.5.

* IFR for Belgium, general population est. 0.30% to 0.62% https://swprs.org/studies-on-covid-19-lethality/ 

* NYC to April 1: 0.4% https://reason.com/2020/06/28/cdc-antibody-studies-confirm-huge-gap-between-covid-19-infections-and-known-cases/

* Missouri April 26: 0.2% (ibid.)

* Utah to May 3: 0.1% (ibid.)

Reason article (June 28) noted "These are just snapshots, and the IFRs in Utah and Missouri may have risen as the epidemic progressed in those states, especially if people infected in May were more vulnerable to the disease." So That 0.1% is likely the best it gets, not the "max" death toll there or anywhere. 

Iceland: situation report - Covid-19 conf. cases 2,165 (+4 today) - conf. deaths: 10. Basic CFR: 0.46%. Unconfirmed infections will be higher. Unconfirmed deaths will also be higher, I bet. I had a look at government stats - small numbers that vary a lot week-to-week (population 341,243 - it's a tiny place.)

https://www.statice.is/publications/experimental-statistics/deaths-ex/

The site notes "It is worth pointing out that the number of deaths for 2020 are preliminary and likely to somewhat underestimate the true number of deaths, mainly due to late arrival of death notices." So there might be more than seen here. Also cause-of-death data is unavailable (another page there explains "Reports on causes of death are usually ready a year after.") But I checked (incomplete) deaths by age as a decent indicator. I'll share this graphic I made to visualize it - deaths of those aged 75 and over are below average as much as above, but it's unusually erratic, and the patterns of just when seem related to the rising cases (peaking a bit before and after peak cases) and perhaps unknown cases in late 2019 (as data in several other places suggests). Being 85+ seems pretty deadly all the time, but note how younger classes (lighter sections and gold rings) seem to die at higher levels in April, just after the peak of known cases, and into May. Also note a rise in elder deaths in July, from quite a low level, accompanying a rise in known cases then - and the deaths fall again as case transmission falls. Possibly coincidence, but likely not.


But this isn't much to clearly suggest additional deaths to go with the additional spread. A  few dozen over-average seem likely - possibly up to about 100 depending on average vs. baseline, and something like 60 as a reasonable high end. But more so than usual, I can see it being zero. 

The Center for Evidence-Based Medicine reported that, for Iceland, "Screening suggests 0.5% are infected;  the correct figure is likely higher due to asymptomatics and many not seeking testing: estimates suggest the real number infected is 1%." The high testing rate was noted, speculating they missed less than half the true infections. That would be a good rate - most places likely miss 3/4 of them or more. 

I have 0.63% now with confirmed infections. Let's say 1% was fair, and 1.5 to even 2% is fair now.

* IFR 0.19% if 1.5% inf. by now (5,119) and still just 10 dead 

* IFR 0.88% if 2% inf. (6,825) and 60 dead

* IFR 1.17% if 1.5% inf. and 60 dead

It's somewhere in that broad range - quite possibly at or near the low end. 0.25% or so may be it. Perhaps the virus did spread widely and less-fatally in Iceland. I suspect not one care home outbreak happened, or their care homes are so tiny you can't tell, so only a worst-case-scenario gets to the 1% range here. 

As such, I suppose the IFR varies well below the 1% seen in hotspots, in the right conditions, and even below the 0.5% end of the CDC's cited range. Down to 0.2% may also be plausible, and even 0.1% - but only in fluke cases (like Utah's early phase). Of course it goes even lower - to about zero - in certain demographics, but then it gets to 30 and 45% in others. But all considered, that 0.6% CDC estimate might be the fairest baseline - places will come out somewhere near, above, or below that line. 

Add Sept. 16: Kendrick: "Sweden, which did not lock down, has had a death rate of 0.0058 percent." But that's not what the Swedes say. Official anti-lockdown government's own website says:

"What is the fatality rate of COVID-19? Globally, it is estimated that 0.5–1 percent of those who are infected with COVID-19 die. ...  this might change as new knowledge becomes available. ...  A study by The Public Health Agency of Sweden estimates the fatality rate in the Stockholm region to 0.6 percent, for all ages. The fatality rate among those 70 years or older is 4.3 percent, whereas it is 0.1 percent among those younger than 70 years."

https://www.folkhalsomyndigheten.se/the-public-health-agency-of-sweden/communicable-disease-control/covid-19/

It seems Kendrick the bogus health data expert crunched out 0.0058 from legitimate data, and conveniently forgot to move the decimal over 2 BEFORE making it a percentage. The report in fact says "Dividing the number of deaths with the number of infections gives an IFR estimate of 0.58% (c.i. 0.37–1.05%)" He re-calculated it himself to be sure, and left the decimal point. 

https://www.folkhalsomyndigheten.se/publicerat-material/publikationsarkiv/t/the-infection-fatality-rate-of-covid-19-in-stockholm-technical-report/