US excess deaths in 2020

I’ve seen coronavirus death totals so far of numbers from 189,000 to somewhere over 200,000. The 189,000 is the only one whose source I remember for sure. It was CNN article quoting Johns Hopkins. I think the 200,000+ number was one I heard on NPR but I’m not sure.

For comparison, the CDC keeps weekly tabs on total deaths from all causes here The last three data points should be ignored because it takes three weeks or so for all the numbers to come in.

So comparing the first 25 weeks of 2020 to the first 25 weeks of 2019, we see 138,191 more deaths in 2020.
But 2019 was a slow year for death. If we compare to the first 25 weeks of 2018, we get only 124,958 more deaths in 2020.

Either of these is a lot of people. But the true effect of the coronavirus can’t be significantly more than these numbers. Can it?

A few confounding factors I can think of:

  • Very sick people who would have died anyway died of Covid a little bit sooner, but would have been counted anyway
  • Millions of people stayed home, so traffic deaths are down, as well as other potentially deadly diseases

I don’t know if these factors are sufficient to explain the discrepancy you’re seeing, but are undoubtedly significant.

I’m not a statistician or a epidemiologist, but I imagine deaths from other causes are way down, for example – traffic and other transportation deaths; flu and other communicable disease deaths; stress-related deaths (this could go either way, I guess – my stress levels are way down).

According to this, as of August 22, there were over 200,000 excess deaths worldwide, which they claim is the COVID undercount.

FWIW the best figuring of excess mortality may take years to figure out. But of course some number of the vulnerable medically fragile individuals who died of COVID-19 were at risk of dying of another trigger in the following six months.

It takes longer than that. Compare, for example, the weekly tab from two months ago to the most recent. Between week 20 and week 28, they added another nearly 12,000 deaths to the first quarter (weeks 1 to 13), including another 448 way back in week 2. I would be very leery of comparing numbers for at least three months and preferably six, because (especially this year) death certificate preparation can take longer than you think. In my state, for example, funeral directors are the ones who submit the death certificate to the state health department, and delayed funerals and backlogs mean they are slower even than usual.

Two things…

  1. It’s clear from the csv you linked to that the most recent weeks do not contain the final numbers, but it is not clear how far back we have to go to have reasonably final numbers. Is week 25 complete? Week 24? We can’t say for sure.

  2. Even if we make the assumption that the data is complete through week 25, week 25 ended on June 21 in 2020 but we’re through 36 weeks of 2020 right now. There were about 125k COVID deaths on that date depending on your data source through 25 weeks. Meanwhile the 2020 cumulative death outpaced the the average of the cumulative deaths from 2014 to 2019 by over 179k over the first 25 weeks. There’s a discrepancy, and it goes in the opposite direction of what was stated in the OP.

Here’s a dataviz. There’s a discrepancy of about 50k deaths through 25 weeks, and that’s 50k deaths under-reported.

Google Photos

Let me be sure I understand you. The claim here is that about 50K people died in the first 25 weeks of the year but don’t show up yet in any publicly available database? That would seem unusual, in this year where we’re seeing data like we’ve never seen before, all the way down to number of tests performed daily county by county.

It’s a pretty big claim. What percent of total deaths, from all causes, are we talking about? I’m also trying to imagine how heavy an overload there must be to cause delays of weeks or months in processing death certificates. Like, a surge of 40 percent I could understand. But a surge of ten percent in workload – I wouldn’t think that would stress the systems too terribly.

Unless it’s just common, in any year, for deaths to take a long time to be processed? I’d buy that as an explanation, as opposed to the workload just being too high this year thanks to covid, but I’d still expect an effort to be made to hurry things along this particular year, given the profound public health implications of measures that are driven (or not, as the case may be) by this kind of data. It would seem both highly negligent and also intentionally foolish to deploy such incredible resources to running hundreds of thousands of PCR tests a day yet being willfully blind to exactly how many citizens are dying.

2018 is a bad year to compare. The 2017-18 flu season was exceptionally bad.

So you are comparing to a year that already had a high excess.

These numbers aren’t straight comparisons because they track flu by season rather than calendar year. But a typical season lately is between 12k and 56k deaths in the US.

It is directly from the data set from the CDC linked in the OP of this thread. It could not be more publicly available.

Let me break it down:

  1. Through the first 25 weeks of 2020 there are 179k more deaths from all causes than the average of the first 25 weeks of the previous six years. That’s from the data linked in the OP.

  2. Through the first 25 weeks of 2020 there were about 125k COVID deaths. Source:

  3. The difference between those two numbers is 54k.

Let me know if you’re having difficulty with anything else.

You said there were ~50K ‘under-reported’ deaths, and you said the data were, or may have been, incomplete. What did you mean by ‘under-reported’? And what did you mean by ‘incomplete’?

Oh, nevermind, I think I get what you are saying. You are suggesting that all those other deaths were actually from covid. I’m guessing they were ‘reported’ as something, though, so I still wonder what it is specifically you mean by ‘under-reported’. Maybe you mean misreported? You’re talking about a ‘discrepancy’, meaning columns don’t match. I’m assuming those deaths that are publicly reported have indeed been placed in some sort of column.

Lots of people have reported that a relative who they know died of Covid-19 was listed as dying from pneumonia, or other causes. Others have reported loved ones dying after experiencing symptoms of Covid-19, but not ever being tested, and having the cause of death listed as a stroke, heart failure, etc., which might have been the immediate cause, but was precipitated by Covid-19. I mean, they were finding people dead in their apartments in NYC at many times the usual rate back at the start, way before there were enough tests.

Also, it wasn’t until a ways in that they began recognizing how much it was causing blood clots, and damage to organs other than the lungs. Probably a lot of deaths were misattributed. Some places have reported pressure to not attribute deaths to Covid-19.

So, here’s some quick arithmetic on the data in the CSV file from the OP. Two columns here, the first with week number and the second with 2019/2020 deaths compared to the average from that same week over the past six years (as Lance_Turbo has done above).

Week Excess
40 1955
41 2772
42 3678
43 3174
44 2431
45 4150
46 3357
47 3756
48 2411
49 3198
50 2984
51 1746
52 2036
1 758
2 561
3 536
4 916
5 918
6 1704
7 1219
8 2010
9 2538
10 2393
11 2312
12 3567
13 8193
14 17487
15 24573
16 23343
17 20778
18 16037
19 13880
20 12080
21 8768
22 6433
23 4503
24 2856

I looked at the weeks going back to 40 of 2019 to see if there were some sort of ‘baseline excess death’ we should take into account. There does appear to be, as you may be able to see from the data.

The last 13 weeks of 2019 ran an excess of 2896 deaths a week, on average, compared to the average of the prior six years.

The first 12 weeks of 2020, before things start to really take off in week 13, run an excess of 1619 deaths a week, on average, compared to the average of the prior six years.

It follows, then, that in the six months or so before the deaths really took off in week 13 of 2020, Americans had been dying at an average of 2283 more per week than they had over the past six years. There does seem to be some sort of baseline of excess death, using this methodology, that we should account for.

In weeks 13 through 24 of 2020 – again, assuming week 24 is ‘complete’, as Lance_Turbo has mentioned – the US ran an excess of about 13,244 deaths a week, compared to the average of the prior six years. That’s 158,931 deaths in total. If you use 2283 per week as a sort of baseline that you’d be expecting to start from, you see a surge of about 131,500 deaths, or about 11,000 a week. Not sure how that would compare to Covid-19 (‘under-reported’ or not) during that same window of time.

I am not, of course, claiming that any of this necessarily reflects reality. But it’s the way I’d approach that data table.

There’s no reason to think that your “baseline excess” would have continued past normal flu season, which I would bet is a primary cause.

It looks like 2019-20 was probably a moderate to fairly bad flu year, but who knows if we will ever know for sure exactly what the split was during flu season. We can be pretty sure that it hasn’t been causing a lot of deaths all summer, though.

Out of curiosity, what do you think could be the cause of some other large quantity of excess deaths.

By the way, baseline excess deaths is a bit of an oxymoron.

Fair, to be sure, but only in the same way that ‘excess’ ever is. I mean, ‘excess’ above what? If what you are comparing something to always gives you an excess, then maybe it’s not an excess after all.

Here’s the full table, with every entry comparing the reported (or ‘under-reported’, as the case may be) numbers from 2019/2020 with the same weeks from prior years. The last column compares 2019/2020 to the average across those years. Not too many negative numbers here, and a whole lot of four-digit ones.

Week 2014 2015 2016 2017 2018 2019 Average
40 5072 3213 1614 1164 457 207 1955
41 5786 3866 3260 966 1513 1241 2772
42 6734 5292 3669 2900 2114 1357 3678
43 6302 5101 3664 2184 1505 285 3174
44 5309 3134 2928 2345 419 448 2431
45 6760 5338 5031 3175 2459 2138 4150
46 6477 3417 4499 2444 2119 1188 3357
47 6606 4435 5001 3069 2077 1350 3756
48 6177 2738 3469 969 705 405 2411
49 6489 3747 3745 2168 1706 1334 3198
50 5810 3036 5816 1681 288 1271 2984
51 4876 1396 4663 728 -2081 896 1746
52 6501 -887 5965 842 -2635 2429 2036
1 6647 -1813 4136 161 -6210 1625 758
2 4795 -726 4890 -565 -7080 2053 561
3 4381 333 3779 -332 -5686 741 536
4 4447 1263 2926 485 -4248 624 916
5 4245 916 3304 -231 -2774 48 918
6 5572 2549 3863 -69 -2141 448 1704
7 4438 1905 2468 -226 -1526 253 1219
8 5579 2553 3348 -67 428 216 2010
9 6731 2673 2530 1149 1713 433 2538
10 6428 3214 2374 1009 1393 -61 2393
11 5764 3284 3084 890 1220 -371 2312
12 7047 5194 3770 1387 2666 1337 3567
13 11509 9358 8487 6664 7449 5692 8193
14 20832 18885 17515 16421 16287 14982 17487
15 27782 25470 24824 23453 23086 22822 24573
16 26285 24413 22872 22951 21983 21553 23343
17 23184 21465 21904 19636 19204 19275 20778
18 18598 16529 16918 15531 14447 14201 16037
19 15152 14801 14473 13488 13368 11995 13880
20 14883 13487 12345 10334 12196 9236 12080
21 10904 9784 9012 8530 8573 5807 8768
22 8594 7716 7046 5501 5712 4030 6433
23 7416 5415 5249 3748 3165 2025 4503
24 5402 4008 3207 2320 1730 471 2856

Oh, I see no reason to doubt that the great bulk of the ‘true’ excess deaths (if that’s something we could ever really know, with a crystal ball to tell us exactly how many people are supposed to die in a given year) are from the disease. I do have concerns about the stories I hear from other countries, not so much the US, about missing half a year, essentially, of things like cancer treatment or whatever else might have helped to forestall death from non-Covid causes in vulnerable people, but I wouldn’t have the faintest guess how many such deaths there might be. And I don’t even want to entertain speculation (of my own) about what effects there might be from lives saved from reduced accidents or taken by drug abuse or whatever. All that seems a bridge too far in any armchair analysis.

Thinking this through a little more, I guess I didn’t fully grasp what you meant at first. I guess you meant that the better way of looking at it is that those weeks spanning week 40 of 2019 to week 12 of 2020 were just especially bad already, even before Covid. Yeah, I guess I get that. Though, six months running is a long time for anything like a flu surge, wouldn’t you think? If you were asking what else might have caused those weeks to be especially deadly, in light of the six years before them, I wouldn’t have a clue.

I’m a little intrigued, though, by the idea that Covid may have been circulating for a little while before we figured out what was going on. Seems absurd, though, to attribute such large, and relatively unusual, numbers to that even so. You’d have to think we’d have figured it out earlier, if that had been the case.

I suppose things might look a little different if we were looking at per capita rather than raw numbers, but I can’t imagine things have changed quite that much over so short a time. I’d also think to look at counting methodologies potentially changing, but we’re talking about raw deaths, right? Hard to imagine we’re simply ‘finding’ more of those than we used to. I dunno.

Not when limited to the first 27 weeks of the year according to the NCHS data that was referenced in the OP. Lance_Turbo has already done yeoman’s work to show that the excess deaths are actually higher than deaths currently attributed to the coronavirus, and I’m pretty sure most of these excess deaths are probably COVID-Related, but attributed to Pneumonia instead. I see no reason in the data to believe that the flu has anything to do with the excess deaths we’re seeing so far.!/vizhome/ExcessDeaths_15997268767130/NCHSData2020throughweek27only

Note the crazy excess of pneumonia deaths for 2020? Pneumonia deaths are pretty stable and if you exclude the outlier year of 2020 there is a standard deviation of only 3,945. When you add in 2020, the number jumps up to 16,631, over four fold.

For shits and grins, drag the slider so the range is weeks 1 through 12 (before the shit really started to hit the fan) and you’ll see that 2020 looks like any other year, both from a total and pneumonia related death standpoint.