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.

Sunday, July 26, 2020

Global "Excess Deaths" in 2020: One Million Already Killed by Covid19?

July 26, 2020
(incomplete)
updates Aug. 2, 6, 31

Even now, and with some degree of under-counting global deaths confirmed as caused by Covid19/SARS2 stand near 650,000 and rising some 4-7,000 each day. It's to the under-counting I now turn, with a start at a clearer view of an enormous number of un-recorded deaths worldwide will combine with what we know to push the pandemic as it stands to the top ranks of recent history's deadliest - likely over one million already - and that's despite existing control measures, and despite being far from over. It's a bit frightening, but as I say, "no panic, ever."

Where I've kept track, mainly, and cited throughout:
https://www.bing.com/covid?vert=graph
https://coronavirus.jhu.edu/map.html

I'll start with "floating" and "flickering" death tolls – numbers I've seen, and that might offer a glimpse of the scale here. Sometimes bulk additions appear in a place's confirmed Covid19 cases or deaths, probably showing the results of some bulk tests of cases past. These seem to mainly be previously unknown nursing home and at-home and even in-the-alley deaths later confirmed as caused by Covid 19, or maybe classed as probable enough to include. Consider NewYork - a "floating" tally of probable deaths seems the diff. between 25,081 confirmed deaths plotted by date on the graph, and the larger 32,409 listed. That's 7,328 deaths that are "floating" but seem fully reported.
https://www.bing.com/covid/local/newyork_unitedstates?vert=graph

The bulk of these are logged in New York city. I presume the state uses the same basic definition as the city: "Cause of death reported as COVID-19 or equivalent, but no positive laboratory test." This should be a pneumonia-like illness and, by the graph, it seems to spread and kill quickly, much like and right alongside COVID19 so it blends right in like an extension of it, even falling the same way under the same control conditions (a standard cold or flu would have fallen of much earlier and more totally). That "probable" call seems likely to explain all, or all but a handful of these deaths.
https://www1.nyc.gov/site/doh/covid/covid-19-data-deaths.page

Smaller but equally hard-hit New Jersey similarly has 13,660 graphed at Bing vs. 15,765 listed there and at the JHU tracker, for 2,105 lodged but "floating" deaths.
https://www.bing.com/covid/local/newjersey_unitedstates?vert=graph

These two states had what seemed by far the deadliest tangle with a raging epidemic, and followed with extra efforts to slow the spread and reduce the death toll, which have been fairly successful. They've also done more than most to gauge the problem, verify the unknowns, etc. As such, their current tallies may be about right. But other states seem to have some reckoning left.

Some states seem to reckoning. 6,374 deaths were just now added for Pennsylvania, another hard-hit state with an aggressive policy against the virus. That nearly doubled the prior toll, leaving a whopping 13,437 deaths now listed dead at Bing (or higher by now, unless it reverts). But this floating tally isn't yet included on the JHU tracker, and just 7,063 remains plotted on the Bing graph of confirmed deaths by date. The massive addition was also not reflected on the state's Covid19 tracking site when I checked, and I didn't find any news story to explain.

But back on April 21, some 201 "probable" deaths had been included, but were removed on review. Is this what was re-added, but massively expanded over time? As we'll see, just this kind of addition is often warranted. Little-known news: Pennsylvania is by far the third-largest admitted death toll, after new York and New Jersey, only the third to cross the 10,000 deaths threshold - publicly.
https://www.heraldstaronline.com/news/local-news/2020/04/pennsylvania-slashes-covid-19-death-toll-by-201/

Similar numbers have briefly appeared in smaller places without yet sticking like they do in Pennsylvania. These I call "flickering," as they appear briefly and then vanish again. I failed to take a screen-grab for any of these, and they blink away almost like errors would. But these wind up looking like real thing people don't fully admit to (ex WHO reports new numbers, they appear on the list, the state complains, numbers vanish again). I've only noted three of any real size:

~4,300 in Florida – A huge number briefly appeared on the JHU tracker the other day, raising the death toll to close to 12,000 (as I thought I had seen - a full doubling - maybe just doubling what i saw for the Netherlands?) or (as I typed it - prefeered!) 9,964. Then it reverted instantly to whatever exactly - 5,632 or lower (risen a bit since). I should have followed that better, but I suspect the right version will come out anyway - they already flickered it. Suggested unreported deaths: 4,332.

~6130 in the Netherlands - both JHU and Bing trackers a few days back briefly showed 12,279 dead for this lockdown success story, roughly doubling the total before reverting to about 6,130 – the Bing counter is now 12 days delayed as well; July 13 is still the last date shown, with a tally of 6,156 dead, a bit higher than listed. Another 1,000 cases have been listed, 137 of them in the last 24 hours. But the graph and the death toll remain on hold for some reason - maybe to work these in.).

~2,800 S. Africa - I also saw back in May South Africa's tally jump by some 3,000 deaths, then quickly revert to the 200 confirmed ones they had. See the first graphic here where I plotted it at 3.5/m (219) on 5/13 to 54/m (~3,100) on 5/21, reverted next day. I estimate the difference from last toll at 2,763 "flickering" deaths. Were they hiding some 90% of a huge death toll? If so, the only evidence I had seen suggests they admitted to it briefly. And evidence considered below suggests these deaths are real, but they weren't confirmed Covid19, only probable. And it seems this flickering segment has grown since, and probably should be added to nearly double South Africa's soaring death toll.

Comparing to "Excess Deaths" 

Looking into it a bit, there seems to be a major overlap with "excess deaths" analysis.  Most places report overall deaths, or ones from natural causes, over regular time spans (weekly in most cases),
and keep track of historical norms fo that same span over some years back. This yields an average number of deaths for that time, and it can be significant if deaths rise above that average level, not even knowing all the causes.

The CDC has a page suggesting how excess deaths statistics might indicate the death toll of Covid19 well above what's known. They even admit these numbers would include other elevated deaths (including lockdown-related: the deaths from cancelled medical procedures, or the increased suicides and domestic violence we've heard about), but they'd also exclude less in other areas (other communicable disease and traffic accidents seem to both be down).
https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm

From that page, 3.5 years of daily, all-causes deaths in the United States - all causes kill randomly year-round. totaling about 51-52,000 per day in the summer, and 58-60,000/day on average in the winter - an unusual high of nearly 68,000 is recorded for the week of January 13, 2018. That's how deadly seasonal illnesses are, and this one's far worse - as shown, peak deaths so far - 79,641 (some 40% above average) - came the week of April 11, with terrible weeks before and after. The deaths only falling like that because people made them fall with unusual and extreme efforts to protect vulnerable and slow the spread. The turn to summer probably has almost no effect - see it rising now as it gets warmer, but people move around more, and many refuse to mask-up, etc. It's year-round flu season, with an extra-contagious and extra-deadly kind of "flu."

Subtracting the average rate from the actual weekly tally (? "projected (weighted)"  - the ones that go taller), the weekly differences, from where I see a difference appearing, are this much above average:
3/14: 8
3/21: 1,252
3/28: 5,842
4/4: 15,843
4/11: 23,383
4/18: 21,723
4/25: 19,397
5/2: 15,067
5/9: 13,327
5/16: 11,357
5/23: 8,896
5/30: 7,204
6/6: 6,446
6/13: 4,235
6/20: 3,812
6/27: 4,288
7/4: 5,256
7/11: 4,295
Total excess deaths: 171,631.
Confirmed Covid19, Bing July 11: 123,862
Likely uncounted Covid19 deaths by July 11: 47,769

There may be other deaths adding to the excess - a relative few - or on balance the other classes might be negative additions (below average, meaning even more Covid ones to explain the full number). As I'll show, the data often suggests "Lockdown" measures prevent even non-Covid deaths on balance, driving the baseline below average. People had less car wrecks - no bars open to die driving back drunk from, Is there less deadly violence when men stay at home, even with their wife and kids? That seems likely, even if domestic battery is on the rise. Other transmissible illnesses will drop sharply. I hear suicides have risen, alcohol and drug abuse, overdoses, as well as other medical deaths caused by fear or cancellation of needed procedures. I'm sure some of each have happened, but confused people have talked about this the most as it it would total millions. It seems from the data that these causes don't add up to much in reality.

It's not economically tenable nor psychologically for most people, etc. But on balance, it seems to me "Lockdown" saved far and away more lives than the relative few it cost. While, as noted, all other deaths combined seem to come out low, the avoided SARS2 deaths are the clear bulk of our collective victory here. It's hard to predict and gauge what they would have been under the many other possible courses of action, but some tragically enormous death tolls have occurred (2.5% of those alive in Bergamo, Italy now are not alive), and more of the same remains possible, while a milder but massive culling is all but sure to continue for the next year and whatever.

So to degree of saved lives isn't clear, and for now we can just note the actual number of un-noted Covid deaths is probably significantly higher than this 47,769. Conservatively, I'd say it's in excess of 50,000 (or as it happens, no less than 1,000 per state on average). Added to the current acknowledged toll of 148,593, that would put us at a minimum of 198,593 deaths so far in the United States.

A recent dispatch from Reuters on South Africa's death toll cites new research claiming that deaths from natural causes (not violence, accidents, etc.) were "59% more than would normally be expected between early May and mid-July, scientists said, suggesting many more people are dying of COVID-19 than shown in official figures. An "excess" of 5,022 deaths by natural causes, about half more than usual, means around 12,068 would usually die in this span. "The council's data showed 17,090 extra deaths, 11,175 were people over the age of 60, a telltale sign of COVID-19, which is overwhelmingly more deadly for older people." Most things are deadlier on that demographic, but it's a major rise above average in the time of COVID19, so yes - the bulk of those 5,022 at least are probably more coronavirus deaths.  Again, this is just in some 2.5 months following national lockdown as soon as here were a few hundred known cases.  Prior to that is another story we'll come back to - the first KNOWN case isn't what matters.

Other natural causes will be down (illnesses at least), after lockdown in late March, said to be a strict one. So there were 5,022 extra deaths left over after some of the Covid deaths cancelled out any below-average levels. We'll come back to that.  Compare Bing's current record for July 14: 4,346 confirmed, just 646 less than the (non-adjusted) suspicious overage of 5,022 (5,940 is cited in the later article - and it's shooting up - now 6,655)

A fascinating bit of work at The Economist recently looked at the excess death numbers is several places into June, including a treasure-trove of infographics to make the point in a sampling of places, including much of Europe, a few bits of the wider world (where such statistics are more rare), and all 51 of the United States + D.C.. Covid19 deaths are shown alongside overall excess for several weekly spans (monthly in some cases), and can be displayed in absolute numbers or proportional (per 100,000 people)

Tracking covid-19 excess deaths across countries 
In many parts of the world, official death tolls undercount the total number of fatalities
July 15, 2020
https://www.economist.com/graphic-detail/2020/07/15/tracking-covid-19-excess-deaths-across-countries

The zero baseline in these graphs "represents the 2015-19 average of deaths from all causes" in this region at this time (including one unusually-deadly flu year, 2017-18). A deadly and spreading virus would appear as a rising mound above the starting level, and what we see in case after case is a rather steep hill rising quickly and massively above average, riding well above the similar-shaped mound of confirmed Covid19 deaths, and sloping back down along with it.  This would include other elevated deaths and exclude less in other areas, as noted ("all causes" - unlike S. Africa example, this includes shooting, car crashes, etc. I think.). But these don't seem to play in so much at the scale that emerges, tending to average together into the baseline we're looking at a huge deviation from.

Shown: S/W.Europe hotspots as they display in absolute numbers - the red and pink numbers shown are just for that date - (the week ending) March 30, the general peak before measures brought deaths back down.


...

Note the UK and Netherlands both had joined Sweden in a managed "herd immunity" approach for a week or so, before these two changed course. That first phase doomed them to a deeper problem than others, smoldering hot and flaring up into a higher long-term death toll than the early days would have suggested. And just as it is with the others, the widespread deaths have gone largely un-counted, at least in mid-late June.

Some cases in Latin America have less data points, but paint a grim picture. The Peru one appeared terrifying when I thought it showed at least 3 times the current death toll there, now past 17,000. As I'll show below, the current tally is about what this shows - the whole red curve was eventually filled in pink - those excess death were eventually classed as majority Covid. The others may still require some revision.

Several studies have shown other transmissible illnesses like influenza and the common cold, fall of sharply once hygiene and distancing advice for SARS2 takes hold, and this increases as lockdowns are imposed. The more-contagious SARS2 is also slowed down at this time, but never stopped - it takes off and kills anyway, just a good bit slower than it would have without these brakes placed on it.

The lag between infections and deaths is why it peaks after lockdown,
and good advice increasingly heeded helps the deaths stay low, sometimes shrinking to average and below even by June. In China, they pushed it down to effective zero deaths by mid-April and it's stayed there.  - but then most only got it well-slowed, then let its leash out so cases skyrocket and deaths started rising again - and more unverified but excessive deaths will be racked up past what those study showed into late June (see US total above).

And note: "Some (U.S.) states take at least a month to process death certificates, which means that the initial numbers they publish for a given week are substantial undercounts, creating the false impression of “negative” excess deaths." Similar issues may play into just when the relevant numbers appear, and maybe in the patterns between points.

I tried to look at the excess vs. Covid deaths visually to estimate the proportional differences for each country shown, and found out I did it poorly in some cases when I checked a few places more carefully. Here I'll include these careful checks. My analysis is unusual, but the few estimates I can check against those floating and flickering tallies bear it out as fairly predictive. Time and again, the data suggests true death tolls are 10-50-100% greater than what's currently reported.

Analysis of several nations, states, and cities
starting at the average line, again covering a multi-year average of that week's deaths, from all causes. Anything might push 2020 deaths above or below that line, from a change in the weather to a halt to gang violence, the arrival of a disease-bearing tick, etc. In most places and most cases, nothing changes enough to much effect an average over several years, where each day comes out averaging like the Angel of Death's agreed tithe of 51,000/week in the USA.

Where I see a steady rise that blends right into the big Covid curve, I start counting - usually confirmed cases appear a little ways in, suggesting those were the low-boiling, unknown early cases. My unusual analysis includes an adjustment from the usual average, wherever a trend that way is clear, and a note that "Lockdown" seems to push it lower yet - clearly it saves lives from Covid19, and on balance even non-Covid ones drop to such a degree it hides the true scale of Covid deaths that happen anyway. This might sound implausible, but the few predictions I can check against floating and flickering tallies bear it out as fairly predictive. How to gauge the below or above average levels is less clear - in some cases I ignore it. But when there's a clear trend before app. Covid deaths, I take a rough average level for it, and use that as the baseline, by adding x/week time number of weeks considered. This is quite imprecise, but it seems the imprecisions balance each other, and it broadly produces predicted numbers quite like those already reported or seen flickering.

The following tries to take down each data point for a total of excess deaths vs. confirmed Covid (C.C.), the difference between them, adjusted for off-average level, and compared with other relevant numbers. I'll start with the ones we can compare against. First, we'll look at Netherlands, Florida, Pennsylvania, and South Africa, where can compare to the floating and flickering tallies discussed above.

Cases we can compare
Netherlands
date | excess | C.C.
3/7 | -234 | 3
3/14 | -106 | 17
3/21 | 463 | 159
3/28 | 1389 | 592
4/4  | 2113 |  - 995
4/11  | 2054 |  971
4/18  | 1423 |  947
4/25 |  1098 |  791
5/2  | 641 |  581
5/9  | 240  | 384
5/16 |  -23  | 240
5/23  | 61  | 142
5/30  | 5  | 134
6/6  | -31  | 57
6/13  | 30  | 46
6/20  | 51  | 31
total: 9,174 excess vs. 5,143 C.C. (Bing 6/21: 6,090 - maybe added to since)
Difference: 3,084
Off-average adjustment: -200/week x15 = +3,000
Likely uncounted Covid19 deaths by 6-21: 6,084
That's should be pretty approximate, buy may be very close - not many more deaths expected since and, again, ~6130 flickered in mid-July.

Florida
date  | excess  | C.C.
3/6  | 344  | 2
3/13  | 254  | 1
3/20  | 221  | 7
3/27  | 332  | 46
4/3  | 435  | 139
4/10  | 539  | 251
4/17  | 482  | 304
4/24  | 671  | 304
5/1  | 435  | 310
5/8  | 466  | 349
5/15  | 446  | 250
5/22  | 556  | 269
5/29  | 331  | 214
6/5  | 316  | 242
6/12  | -52  | 237
total: 5,776 excess vs. 2,925 C.C. (Bing 6-13 total: 2,925)
difference: 2,852 
Off-average adjustment: +120/week xfirst 3 weeks? = subtract 360 (bad flu season, falls off sharply in late March). Other causes may fall, but let's skip that.
Likely uncounted Covid19 deaths by June 13: 2,492

Suggested current half-reported: ~6,000. It has been a bad month and a half. In fact ...
“Part of deaths may be due to non-Covid conditions, things like heart attacks or cancer or stroke and those patients who were not able to receive care during the pandemic either because of fear of catching COVID in the hospital setting or hospitals that had to postpone care in order to create capacity to meet the need of COVID-19 cases,” Dr. Tsai said. “You have to understand both to understand the true cost of the pandemic.”
A look at the past six months in Florida shows why health experts are increasingly concerned about death toll numbers here. Between January to June 30, Florida reported 8,671 excess deaths, that’s more than double the 3,650 deaths the state attributed to COVID-19.
https://www.abcactionnews.com/rebound/excess-deaths-raise-concerns-about-pandemics-death-toll-in-florida
Many excess deaths are in Jan. and Feb. - probably before COVID19 arrived - maybe a bad flu year?
Yes, the normal seasonal influenza was causing hospital visits and deaths above average in late 2019, worse into January and February, then dropping sharply around week 13 (late March), probably in part die to Covid19 measures enacted around then by most states.
http://www.floridahealth.gov/diseases-and-conditions/influenza/index.html

6/30: 8,671 excess vs. 3,505 C.C. (Bing). difference: 5,166. I suggest the bulk of these - more than 5,000 - are yet more covid ones. Considering a possible lockdown-lowered deaths baseline, it might be more than you'd think. Nearly a month later, it might be that near-6,000 number I saw flickering.
Likely uncounted Covid19 deaths 
in Florida, by June 30: 5,100

Pennsylvania
date excess C.C.
3/27  | -35 | 32
4/3  |  247  | 102
4/10  | 698  | 355
4/17  | 689  | 345
4/24  | 1122  | 700
5/1  | 930  | 881
5/8  | 892  | 1252
5/15  | 706  | 534
5/22  | 383  | 893
5/29  | 307  | 440
6/5  | 132  | 393
6/12  | -45  |  280
total 5328 - 6207 C.C. (6-13 Bing 6211)
Difference: -879 (less than average!)
Off-average adjustment: -400/w x 11 = add 4,400
Likely uncounted Covid19 deaths as of 6-30:  3,521
Sound crazy? Actually added in late July: 6,374 
noting excess deaths actually dropped, for whatever cause in February, they must have fallen massively for the mountain of Covid deaths to only be half visible above that crater's edge. That leaves even me wondering, but I see the same thing in other places, just to a lesser scale.

S. Africa
The graph is hard to read, with sporadic up-and-down spikes before lockdown, probably reflectind a cycle of delayed reports adding up differently and best averaged. This early high baseline runs back to a possible unexplained high near 1,000 above on January 6. (That's not plotted, but I have to include it.) I get an average of 213 above normal, over 10 weeks prior to lockdown (up to 3-23). There are 2,163 excess deaths, no confirmed Covid. But it could be 2,060 of those (my guess) were from SARS2 getting some early start before it was even being looked for.  
"President Cyril Ramaphosa implemented a tough lockdown at the end of March, shutting shops, ordering people to stay home and sending the army on to the streets to enforce it, back when South Africa had only 400 recorded cases." I had seen the flicker of ~3,000 back in late May, sometime after this when it was at around 20,000 confirmed infections and ~360 confirmed dead. It seems lockdown brought deaths amazingly low, then it goes bad - too erratic all around. Around 700 below/week is achieved as baseline for 14 weeks, but "a surge in poverty and unemployment in a country that already had too much of both spurred the government to lift restrictions well before the peak of infections." Cases have been soaring though June and July, and deaths are climbing fast as well. Counting from just before lockdown:

date excess C.C.
3/23 25 0
3/30 -539 5
4/6 -711 8
4/13 -484 14
4/20 -690 31
4/27 -1,463 35
5/4 -1156 55
5/11 -1038 58
5/18 -754 106
5/25 -810 212
6/1 -329 231
6/8 -758 407
6/15 -138 463
6/22 763 477
6/29 1713 555
Total: -6,369 excess vs. 2,657 C.C. (Bing 5/25: 2,529)
Difference: -3,712
Adjust for off-average: -700/week x14 = +9,800 
Adjusted post-lockdown total: 6,088
Adding the 2,060 pre-lockdown probable-seeming deaths =
Likely uncounted Covid19 deaths: 8,148

That may be high, but compares decently with other readings suggesting something well over 5,000 just the span where my method predicts about 6,000. 

Other places with big differences
I didn't try to plot out all the other places graphed in that study at The Economist. Many seem to have little disparity between the pink shape and the red curve, suggesting they're more full reported. These include: Belgium, France, Switzerland, Austria Denmark, Norway, and Gemany, which for one might be worth a close look. Portugal may also be worth a look, but I skipped it for now. I also skipped a few of the wider-world examples, at least for now; These include: Chile, Rio De Janeiro, and I guess that's it.

Britain/(UK)
date  | excess  | C.C.
3/12  | -188  | 5
3/19  | -207  | 114
3/26  | 1113  | 610
4/2  | 7122  | 3812
4/9  | 8979  | 6899
4/16  | 12731  | 9509
4/23  | 12358  | 9024
4/30  | 8464  | 6685
5/7  | 3842 4429
5/14  | 4704 4220
5/21  | 2503 2872
5/28  | 1456 2002
6/4  | 1189 1697 
6/11  | 485 1204
6/18  | 581 849
6/25  | -339 653
Total: 52,623 excess vs. 54,579 C.C.
Difference: 1,956
Adjust for off-average: -750/week x15 = 11,250
Adjusted difference: 13,206
Bing 6-25: 43,414 - an incomplete tally, short by 11,165.
Those are similar numbers, but for different thing: another 11,165 have been confirmed (C.C. total), and yet another 13,000 are suggested as dying from presumably the same thing. 
Total possible addition to current 6-30 tally: 
24,371 (a 56% increase)


Italy
date excess C.C.
2/24 -437 vs. 10
3/2 -184 69
3/9 1913 552
3/16 5638 1857
3/23  9494 4238
3/30 9694 5462
4/6 6745 4521
4/13 5373 3834
4/20 3352 3478
4/27 1829 2654
5/4 690 1921 
5/11 216 1580
5/18 -53 1247
5/25 -782 780
Total: 43,488 excess vs. 32,203 C.C. (Bing 5/25: 23,877)
Difference: 11,285
Adjust for off-average: 1000 below each week x13 
+13,000 = 24,285
Total: 56,488
Bing 6-25: 34,678
Likely uncounted Covid19 deaths: 
21,810 (a 62.6% increase)

Spain 
date excess vs. C.C.
2/24 -604 0
3/2 -439 1
3/9 -189 65
3/16 1482 763
3/23 6100 3354
3/30 11498 5862
4/6 10675 5319
4/13 7099 4250
4/20 4136 3005
4/27 1974 1967
5/4 1110 1328
5/11 391 898
5/18 -158 477
5/25 236 388
6/1 129 235
6/8 -495 137
6/15 -270 83
6/22 -51 22
6/29 234 27
Total: 42,858 excess vs. 28,115 C.C. (Bing 6-29: 28,346)
Difference: 14,743
Adjust for off-average: unclear, ignored
Likely uncounted Covid19 deaths: 
14,743 (a 52% increase)

Noting: Spain seems to have an early outbreak of something deadly, starting off below average, rising quickly to 57 over Jan. 13 (first data point),
increasing steadily to Feb. 3, then dropping just a bit before confirmed Covid deaths start rising sharply - out of nowhere?  The same but milder appears in Portugal, bleeding over already in February, but not causing as much damage. "The virus was first confirmed to have spread to Spain on 31 January 2020, when a German tourist tested positive for SARS-CoV-2 in La Gomera, Canary Islands." Considering The South Africa case, someone else may have brought it earlier, around Christmas. The flu news didn't give any reason to suspect such heavy fatalities.
http://outbreaknewstoday.com/spain-flu-epidemic-wave-of-the-2019-2020-season-has-officially-begun-33028/
However, the sharp drop in February would need explaining - was there some early lockdown or quarantine in the vulnerable community, some days to weeks before that drop? Most likely this is a deadly flu, with deaths falling in February like most places, and soon replaced with Coronavirus widening in late February. As such, I didn't count those early highs, but I can still see either being the case, and that might be another 1,000 or so unknown Covid deaths. 

Sweden:
date excess vs. C.C.
3/16 -93 v 8
3/23 83 73
3/30 278 251
4/6 628 532
4/13 833 667
4/20 703 628
4/27 543 555
5/4 590 544
5/11 516 479
5/18 357 354
5/25 288 331
6/1 135 279
6/8 166 243
6/15 166 219
6/22 36 179
total: 4396 excess vs. 5342 C.C. (Bing 6-23: 5276)
Difference: 946
Adjust for off-average start: about 150/week x14
Likely uncounted Covid19 deaths to 6-23: 
3,046 (a 58% increase)

Ecuador
date excess vs. C.C.
2-28 47 0
3-30 723 18
4-29 3219 192
5-30 811 555
6-30 525 273
tot 5325 excess vs. 1038 C.C. -
Difference: 4287
No adjustment for off-average
Likely total by that: 5325
Bing 6-30: 4527
Likely uncounted Covid19 deaths: 
798 (an 18% increase)

Peru
date excess vs. C.C.
3/30 120 vs. 7
4/29 981 238
5/30 3478 780
6/30 3877 1207
tot 8456 excess vs. 2232 C.C.
Difference: 6224
Adjust for off-average: -200/month x4 months = -800
Likely uncounted Covid19 deaths then: 5,424
But the Bing counter shows now for 6-30: 9,677, adding 7,445 to the 2,232 graphed here.
Those calculated and more have been included - new deaths are added, and in 7-22 another batch of nearly 4,000 was added. The current 17,843 is likely right. Good - that's high enough for a death toll that's still far from complete.

Mexico City
date excess vs. C.C.
3/14 -9 0
3/21 -45 1
3/28 51 6
4/4 23 12
4/11 208 37
4/18 442 127
4/25 1094 126
5/2 2129 163
5/9 2722 324
5/16 2796 536
5/23 2960 631
5/30 2486 695
6/6 2451 1071
6/13 2147 824
6/20 1857 898
6/27 1392 909
tot 22,704 excess vs. 6,360 C.C.  (Bing plots 6,360 only on 7/2)
Difference: 16,344
Adjust for off-average: seems to start low, but let's skip it
Likely uncounted Covid19 deaths by 6/27: 
16,344 (257% more than reported)

Moscow 
3/30 -86 0
4/29 401 143
5/30 1305 421
6/29 836 308
tot 2,456 excess vs. 872 C.C.
Difference: 1,584
Adjust for off-average: 50 below each month x3 = +150
Total: 1,734
Bing 6-30: 3,761 - seems to include this and 2,000 more.
Likely uncounted Covid19 deaths: none apparent, now. But Moscow city has 3,761 dead so far, and I bet they're still far short of any herd immunity threshold.

Istanbul
3/9 -57 0
3/16 88 0
3/23 283 22
3/30 444 85
4/6 818 256
4/13 782 339
4/20 629 428
4/27 481 367
5/4 374 264
5/11 287 187
Total: 4,129 excess vs. 1,948 C.C.
Difference: 2,181
Adjust for off-average: unclear, ignored
I don't have another source handy for Istanbul to compare (Bing doesn't list it)
Likely uncounted Covid19 deaths by 5/11: 
2,181 - more than twice what's acknowledged.

Jakarta
(note: Indonesia was one of a few nation that reportedly failed to implement a lockdown policy when others were)
date | excess | C.C.
2/28 | -7 | 0
3/30 | 311 | 19
4/29 | 369 | 69
5/30 | 340 | 31
Total: 1,013 excess vs. 119 C.C.
Difference: 894 (750% above what was reported!)
Adjust for off-average: unclear, ignored
Bing currently shows 516 dead by 5/30 - much higher than graphed, but still incomplete.
https://www.bing.com/covid/local/jakarta_indonesia?vert=graph
Likely uncounted Covid19 deaths by 5/30: 
378 (318% of what's reported)

In the United States
Besides Florida and Pennsylvania, I analyzed five populous states with what looked like big gaps between confirmed Covid and overall excess deaths: NY and NJ get another review suggesting even their "floating" tallies probably miss a lot of deaths, and California, Texas, and Illinois come out two steps behind for lacking even that floating tally of probable deaths. Most states seem to have a similar issue at play, but most of them on a smaller scale, and/or on smaller populations. 

New York
date excess vs. C.C.
2/28 | 88 | 0
3/6 | -6 | 0
3/13 | 75 | 3
3/20 | 460 | 75
3/27 | 2257 | 914
4/3 | 6726 | 3532
4/10 | 9277 | 6588
4/17 | 7157 | 5117
24/4 | 4493 | 5242
5/1 | 2591 | 2513
5/8 | 2033 | 2737
5/15 | 1277 | 1246
5/22 | 641 | 920
5/29 | 525 | 671
6/5 | 270 | 550
6/12 | 86 | 428
total 37,950 excess vs. 30,536 C.C. 
Bing 6-12: just 24527 - so the floating tally (a bit smaller then) is included in 30,536 C.C.
Difference: 7,414
Adjust for off-average start: -70/week x15 = +1050  
Likely uncounted Covid19 deaths to 6-12: 
8,464


California 
date excess vs. C.C.
1/17 28 vs. 0 
1/24 213 0
1/31 27 0
2/7 306 1
2/14 -4 0 
2/21 271 1
2/28 59 0 
3/6 191 2
3/13 -11 3
3/20 417 23
3/27 485 94
4/3 725 199
4/10 895 309
4/17 1057 515
4/24 784 543
5/1 646 495
5/8 728 504
5/15 774 519
5/22 561 527
5/29 553 406
6/5 394 463
6/12 147 451
total: 9246 vs. 5055 (5064 6-13)
Likely uncounted Covid19 deaths to 6-12: 
4,191



Texas
date excess vs. C.C.
1/31 35 0
2/7 144 0
2/14 244 0
2/21 157 0
2/28 105 0
3/6 122 0
3/13 176 0
3/20 303 5
3/27 386 25
4/3 369 77 
4/10 433 147
4/17 527 199
4/24 631 169
5/1 440 224
5/8 544 202
5/15 516 256
5/22 439 199
5/29 295 142
6/5 142 169
6/12 -38 141
total 5970 v 1955 (6-13: 1957)
Likely uncounted Covid19 deaths to 6-12: 
4,015


New Jersey
date excess vs. C.C.
2/28 58 0
3/6 4 0
3/13 -20 1
3/20 108 15
3/27 673 124
4/3 1981 697
4/10 3252 1346
4/17 3183 1888
4/24 2395 1795
5/1 1494 1876
5/8 1183 1373
5/15 912 1134
5/22 554 831
5/29 340 553
6/5 141 472
6/12 -253 482 
total 16,006 excess vs. 12,587 C.C. (Bing 6-13: 12,621)
Difference: 3,419
Adjust for off-average start: -60/week x19 = +1,140
Adjusted difference: 4,559
Minus ~2,000 already "floating"
Likely uncounted Covid19 deaths to 6-12: 
2,560


Illinois
date excess vs. C.C.
1/31 -1 0
2/7 71 0
2/14 15 0
2/21 57 0
2/28 108 0
3/6 110 0
3/13 97 0
3/20 179 6
3/27 211 41
4/3 557 197
4/10 825 433
4/17 906 582
4/24 999 616
5/1 926 679
5/8 1002 793
5/15 850 780
5/22 635 663
5/29 536 540
6/5 479 533
6/12 272 426
total 8,834 excess vs. 6,289 C.C. (Bing 6-13: 6,289)
Difference: 2,545
Adjust for off-average start: unclear, skipped
Likely uncounted Covid19 deaths to 6-12: 
2,545


Total and Context
United States total - skipping the 7 big state estimates and using the national weekly excesses outlined near the top, adjusted for wide range for up to 40% more than that, we can start the total with:
  • United States: 50,000-70,000
  • Netherlands: 6,130
  • South Africa: 8,148
  • United Kingdom: 24,371
  • Italy: 21,810
  • Spain: 14743
  • Sweden: 3,046
  • Ecuador: 798
  • Mex. City: 16,344
  • Istanbul: 2,181
  • Jakarta: 378

Likely uncounted Covid19 deaths to June 30 (at the latest),
from just 8 nations and 3 major cities:
133,206 to 153,206

That's noting what seems to be much fewer exclusions/better accounting in Moscow, in Peru, and in several European nations. But this is not including Canada, Brazil, India, or Pakistan, missing most of Africa, Eastern Europe, the rest of Indonesia, Russia, Mexico, the rest of Latin America, the whole of Central Asia ... I doubt we're missing very many deaths in China, but the trends leading to what we see above - suggestions of 50-100-300% more deaths than reported - must be a global problem. It will vary in size, but I'm sure that, on global average, there are at least 30% more deaths than known. That's only 195,000 more than listed, and I just showed up to 153,000 from 8 nations and 3 cities alone. On a worldwide total of nearly 650,000, that would make at least 845,000 dead so far - probably well over that, and quite likely past one million. 

... more on context f/c ...

More on Bulk Death Additions
Peru, as noted, seems to have few uncounted deaths - some deaths added on July 22 saw to that. 3,876 added - subtracting 195 for daily average new deaths = ~3,681 added deaths.
I estimated 13,880 Covid19 deaths by 6/30 (adding confirmed to excess, w/rough adjustment for low baseline), and the Bing counter now lists (9677+3681=)13,358 - my estimate was over by just 522.

Pennsylvania: As it's not shown on the graph and I didn't note the exact date the new toll appeared, I checked daily fatalities at the Bing tracker - they're still ranging from 4,000-6,000 most days, but 13,778 were added on July 23 - about twice the usual, or over by about Pennsylvania's huge addition.  I think that's when it was added, a day after Peru's. It was about this same time Florida and the Netherlands flickered their own tallies of about the same size.

It's noteworthy that my estimates don't seem high - Peru was by a bit, but Pennsylvania I under-guessed by nearly 3,000, or some 50%. I suppose it varies a lot, is hard to predict, but I seem to be doing pretty well so far - S. Africa, Florida, and the Netherlands all had estimates quite close to what I've seen flickering.

Chile has a similar issue with Peru, a batch of deaths added 7/17. So I went back to its chart:
Date Excess C.C.
1/6 238 0
1/13 244 0
1/20 154 0
1/27 113 0
2/10 103 0
2/17 94 0
2/24 -58 0
3/2 118 0
3/9 164 0
3/16 115 0
3/23 151 2
3/30 162 10
4/6 181 31
4/13 111 49
4/20 199 55
4/27 136 60
5/4 88 68
5/11 410 60
5/18 467 174
5/25 852 297
6/1 1279 382
6/8 1621 1095
6/15 1678 1100
total: 8,620 excess vs. 3,383 C.C.
Difference: 5,237
Off-average adjustment: unclear, above average may all be Covid19, dying in numbers before January even started. 
Likely uncounted Covid19 deaths as of 6/15:  5,237
Bing tracker: 6/15: 3,362,
https://www.bing.com/covid/local/chile?vert=graph
7/17 1,057 deaths added, including daily average 91. Minus that, ~966 prior deaths added.
Projected uncounted deaths after 6/15: much lower – the week after some adds on June 7, the counted deaths are so much higher (200/day, up from 100), with no rise seen, it seems they're counting differently, finding more, close to the gap seen in the Economist graph, where the scary rise reflects the start of this counting. That lower proportion might apply through the sharp rise in cases and deaths since – now at 9,240 confirmed deaths.
The 7/17 add of 966 should include some from after 6/15, but let's just subtract them all for a low-end estimate: at least 4,271 deaths to 6/15 left uncounted. 

Others I've noted with bulk additions at about this time are mostly done between July 16 and 18. These just appear at the Bing tracker - I don't have excess deaths data handy, and some may not bother producing it, as they've been at war, etc.

* Kyrgyzstan: 7/18 some 717 added (jump 173 to 900, minus 10 expected daily add), and more deaths are counted since then - now 1,347. Add 717 = 414.45%
https://www.bing.com/covid/local/kyrgyzstan?vert=graph
* (Kazakhstan plans to start adding their mysterious "pneumonia" deaths in August - at least the new ones. Some 3,000+ prior deaths, perhaps not. They've been counting a lot more since then, with bulk updates every few days, now at 793)

* Yemen: jumps 364-433 on 7/16 - minus 10 daily average = 59 added (15.8%)
https://www.bing.com/covid/local/yemen?vert=graph

* Zambia 42-109 on 7/18 - not even counting for several weeks = 67 added (159.5%). Now counting,  142 confirmed on 7/29.
https://www.bing.com/covid/local/zambia?vert=graph

* Liberia: 51-68 on July 17, (less than 1/day before and after) = 17 added (33.3%)
https://www.bing.com/covid/local/liberia?vert=graph

* Libya: 38-46 on 7/17, 1/day average = 7 added (18.4%)
https://www.bing.com/covid/local/libya?vert=graph

* Benin: 23-31 over July 16-17, minus 1 (less than 1/day before and after) = 7 added (30.4%)
https://www.bing.com/covid/local/benin?vert=graph

* Malawi: 19-31 on 7/10,  less than 1/day before and after = 12 added (63.2%)
https://www.bing.com/covid/local/malawi?vert=graph

* Zimbabwe: 13-23 on July 17, less than 1/day before and after = 10 added (77%)
https://www.bing.com/covid/local/zimbabwe?vert=graph

* Syria 16-35 in 2 steps on 7/19 and 7/23, less than 1/day = 19 added (118.75%)
https://www.bing.com/covid/local/syria?vert=graph

* Buenos Aires, Argentina, added 304 deaths on July 5 (jump 400 to 704 - 76%), though higher numbers had been toggling for days - possible coincidence, but more deaths are counted each day since then (now at 1,397)
https://www.bing.com/covid/local/cityofbuenosaires_argentina?vert=graph

Chile and probably Kazakhstan, but none of these others, adds to the tally of likely uncounted Covid deaths - the others are known of from being added already, and fitting the same pattern of large percentages left out before and just now being properly included.
Add:
Chile 4,271
Kazakhstan: 3,300

Update August 2 - In visual form: I tried to cram all this and more into one infographic, for my coverage of span 10, July 30: "unconfirmed, excess, and pneumonia deaths: the true toll starts to emerge"




Update Aug. 6: 
https://www.nytimes.com/interactive/2020/04/21/world/coronavirus-missing-deaths.html
A NYT study similar to the Economists but broader, does some of the match, coming up with similar likely undercounts, but done simply counting from expected-average as zero, with no effort to adjust the baseline. Understandable - it's a speculative exercise. But I like how my efforts panned out, and the article includes this:
"Today’s rise in all-cause mortality takes place under conditions of extraordinary measures, such as social distancing, lockdowns, closed borders and increased medical care, at least some which have positive impacts,” said Vladimir Shkolnikov, a demographer at the Max Planck Institute for Demographic Research. “It is likely that without these measures, the current death toll would be even higher."
It's not likely but obvious that the Covid-inclusive death toll would be well higher. I think he means to note what I do - other causes also fall on balance, so these measures have a "positive impact" on excess deaths.

https://sciencebasedmedicine.org/COVID-19-and-excess-deaths/
"As a side note, we are seeing fewer deaths from homicides and car accidents during shutdowns, and this is obscuring some of the rise in excess deaths from other causes."

This and lessened deaths by the flu, etc. - an obvious and noted effect - and whatever else seem to add up to a pretty big side-note, actually, more than countering whatever deaths lockdown policies did wind up precipitating.
 
https://www.medicalnewstoday.com/articles/35-percent-of-excess-deaths-from-pandemic-not-caused-by-covid-19
This message of this article is muddled - (were 35% of U.S. excess deaths NOT caused by Covid 19, or did deaths "not directly" caused by it go uncounted? (does "not linked" mean not caused, or that the cause went unnoted / "not linked" in peoples' minds?)  I guess they mean the latter, but cite a study that's heavy on worries about lockdown causing many of the deaths:
"The study also found that excess deaths not linked to COVID-19 rose significantly in states that had the largest outbreaks of the disease during the virus’s peak in early April. These included Massachusetts, Michigan, New Jersey, New York, and Pennsylvania. For example, in these states, there were 96% more diabetes-related deaths than experts predicted. For heart disease, the figure was 89%; for Alzheimer’s disease, it was 64%; and for stroke, it was 35%." 

I'm not sure about how Alzheimer's kills, the rest are cardio-vascular, just the kind of deaths Covid-19 can cause with its attack on the circulatory system (see here for example, and the same article mentions it). They were all way above average, right where and when the virus was at its busiest work. And it seems Pennsylvania has included some 6,400 of these deaths "not linked to COVID-19" as probable COVID-19.

As for the stated worries that these deaths by stroke, etc. may be caused by limited hospital capacity and/or fear the hospital, the piece at sciencebasedmedicine.org includes this: 

"For example, people have been fearful of going to the ER.  Hospitals have been admitting fewer strokes and heart attacks overall. There is no reason to think there is a real drop in the incidence of stroke, and so it is likely that people are just staying away. This will have an obvious negative effect on the outcome of these serious conditions. Further, we may be seeing the effect of hospital and other health care resources being overwhelmed. When the ICUs are full of COVID-19 patients, patients with other serious but unrelated medical problems may not get the care they would have outside the pandemic."

I'm sure such factors have been central to at least a few times, but I don't see it being significant. Someone is HAVING a heart attack that proves fatal, - and NOT after getting the virus - and they stay home for fear of getting the virus? Maybe if it's not clear how bad it is ("maybe it heartburn")... As for low capacity/people turned away: I don't see that happening much if they come in with a life-threatening crisis that could be resolved with known procedures. They would turn away people wanting boob jobs or anything optional, but only in the darkest hours of the worst crises in Wuhan, Bergamo, or Guyaquil would the dying be flat-out refused. Most likely, 90-99% of the above-average deaths in these categories were triggered by unknown Covid-19 (and among the expected ones - proportionally, a few may have got Covid just before their date with destiny, but let's ignore it). Overwhelmed hospitals played the part they did (varying) in the hundreds of thousands of excess deaths all over the world, close to 100% of whom had Covid-19 as the cause of death (or the key co-morbidity, if you insist).

And just noticing: Texas made a revision - "more than 12% of the state's death tally was unreported by state health officials before Monday. The Texas Department of State Health Services is now counting deaths marked on death certificates as caused by COVID-19." Previously, it had to pass some other hurdles as well. https://www.houstonpublicmedia.org/articles/news/health-science/coronavirus/2020/07/27/378648/texas-count-of-coronavirus-deaths-jumps-12-after-officials-change-the-way-they-tally-covid-19-fatalities/ Bing tracker 7/25 sees a big spike of 675 deaths added, to the 150-200 usual for a day. The following days see a notably higher number of deaths counted, or added in batches, decreasing from 30--250 back to 200/day. Above, I found about 4,000 likely deaths were left uncounted - about 200% of what they had confirmed. If so, they're still unlisted, and this is a smaller, narrower class of actually confirmed deaths.

Update 8/31: found the cited study on increased death in NY, PA, etc. JAMA:
"Between March 1, 2020, and April 25, 2020, a total of 505,059 deaths were reported in the US; 87,001 (95% CI, 86,578-87,423) were excess deaths, of which 56,246 (65%) were attributed to COVID-19. ... 
The 5 states with the most COVID-19 deaths experienced large proportional increases in deaths due to nonrespiratory underlying causes, including diabetes (96%), heart diseases (89%), Alzheimer disease (64%), and cerebrovascular diseases (35%) (Figure)." 


"New York City experienced the largest increases in nonrespiratory deaths, notably those due to heart disease (398%) and diabetes (356%)." Of course they also suffered the sharpest and deadliest outbreak of the virus. Coincidence?

No comments:

Post a Comment

Comments welcome. Stay civil and on or near-topic. If you're at all stumped about how to comment, please see this post.