Sunday, May 24, 2020

Inflated Death Statistics

There have been claims over the last few weeks of C19 deaths being inflated, of everything being shoved into that category for political purposes. These are based on isolated reports, as far as I can tell, of state or county officials insisting that death certificates be filled out in particular ways that are not justified, and people extrapolating that for "what might be happening" over the entire country. Some are going whole hog and insisting that this must be what is happening. Also, some reporting agencies are showing a time lag in number of deaths, trying to get the numbers as verified and exact as possible before reporting them. Some folks are seizing on that lag time to insist that the deaths are fewer than is being reported to the public.

The opposite is more likely true. (I just used that about GMO foods, too, and highlighted a Chesterton paradox about talking to strangers. Maybe that should have been the name for this blog right from the start.  Another possible choice at the time was "Do I Have To Pull This Car Over?") You may recall a post of mine in early April "Errors In Both Directions." This will all take a while to sort out, but at the moment it seems that there has been an unnaturally high spike in combined influenza/pneumonia recorded deaths beginning about 10 weeks ago.  Those are usually bout 6-7% of all deaths but spiked up as high as 16% of all deaths mid-April.

Most of that seems to be an increase in pneumonia, not influenza deaths when one breaks it down further.  It is also true across the country, every county, not just a few hotspots. The graph is interactive and you can get more information about their methods at the CDC's Fluview site. I can't tell you that any of those excess pneumonia deaths are actually C19 - not "from CoVid," "with CoVid," or "in the general vicinity of a guy who once thought he might have CoVid." But I also don't have any other plausible explanation why those deaths would start spiking early March with a peak in mid-April. This is all in the context of shutting down and less contagious contact overall, which we would presume would reduce all contagious illnesses, including the common colds.

To my eyes, the right side of that graph is res ipsa loquitur. Something out-of-the-ordinary started driving up flu and pneu a little over two months ago.

Update:  I just heard of a possible explanation: false negative C19 tests. This secondhand from an ER nurse.  They have had people come in whose symptoms look like CoVid but test negative.  At that point the default diagnosis is pneumonia.

8 comments:

  1. FWIW I believe you are correct. COVID was much more prevalent in the US than we thought in the early months of 2020. I don't buy the 'I had it around Thanksgiving (2019)' stories but nobody was looking for it here due to a combination of the CDC botching the tests and nobody caring until Italy got slammed.

    ReplyDelete
  2. Correct me if I'm wrong, but isn't the worst effect of the COVID-19 disease the viral pneumonia?

    There have been some reports of people succumbing to cardiac events, but this was always after the pneumonia.

    I'm thinking that the giant spike of pneumonia includes most COVID-19 diagnoses, rather than being something in addition.

    ReplyDelete
  3. @Christoper B - I agree. I'd also like to point out that in late February/early March the local school district was experiencing high absenteeism rates to the point of discussing temporarily closing the schools. I don't think this was Covid19, but it deserves mentioning that this wasn't a mild flu season in every locale.



    ReplyDelete
  4. I think this was specifically CoVid excluded, but I will try to suss it out tomorrow. I may have missed a trick here. There are 160,000 "mixed deaths," but I admit I haven't given great thought to exactly what that means. I think it means, first of all, that at least a few of those might be added to the C19 total rather than subtracted. See also my update in the post.

    ReplyDelete
  5. Hi AVI,

    I have been following your blog for years and always enjoy reading your take on things. Your posts are always thoughtful.

    I tend to believe that the statistics regarding COVID-19 are being juiced but, I admit, I have been heavily influenced by William M. Briggs. Some of my background overlaps with Briggs' and I find his critiques of statistical models to ring true. Here is a link to an article that he wrote in April directly making the case that the COVID-19 statistics have been juiced:

    https://wmbriggs.com/post/30429/

    Even though that is a month old he hasn't walked it back.

    Maybe you could get BS King to take a look at his work? Some of his articles should be very interesting to her.

    Thank you!

    Glenn

    ReplyDelete
  6. And just to drop as a data point, we know that as of the 2nd week of May the number in the press for Colorado COVID-19 deaths was 30% higher than the number of deaths recorded with CODID-19 test or diagnosis as the cause or contributing cause of death. This resulted from coroners making a stink about deaths from accident and overdose (of people who were found in autopsy to have been SARS-nCOV-2 positive) being included in the COVID-19 deaths.

    Not anecdote, not isolated cases. A quarter of the ~1150 deaths the state had on the website as COVID-19 deaths, the state knew to have died from non-COVID-19 causes, and Governor Polis in a press conference suggested that the number of "maybe it is, maybe it isn't" amongst those was in single digits.

    Some of the articles on this go into justification of why it is important to record "died with" cases, so I'm not suggesting that malfeasance is involved, and the director of the health department is very apologetic that we all misunderstood what the headline numbers were showing.

    ReplyDelete
  7. @Glenn - good to hear from you! I've been passing things along to AVI as I see them, though I've been spotty as I have my own somewhat COVID related issue to deal with - migraines. According to one of the tracking apps I use migraines have gone up with their users 25-50% since March. Hardly scientific, but it's certainly been my experience. Makes it hard to focus on anything for too long.

    I had passed this graph along and had initially thought that pneumonia deaths excluded COVID deaths, but I looked further based on Unknown's comment above and saw I was wrong. Mea culpa. To make up for my haste, I took another look to see if I could break this down a little further.

    The CDC provides a total of all pneumonia deaths and a total of those who died of pneumonia and COVID, I'm assuming subtracting pneumonia + COVID from the total pneumonia gives pneumonia only. The quick math I'm doing shows those are still elevated for this time of year, at least up through mid-April with the incomplete data. In March and April we had 3 weeks when we surpassed 4,500 pneumonia deaths without any COVID diagnosis. Downloading the data for the last 5 years, we never see those totals outside of Jan/Feb. Late March/April median looks like it's about 3,500 deaths/week....so we're running 1,000 or so extra a week with the data not fully in yet. The CDC website says it takes 8 weeks to get 75% of the data filed, just to give a picture here. Of course with all the scrutiny on respiratory illnesses people could be filing these faster, so who knows where this will go.

    For those interested in the historic data, the fluview website AVI linked to has an option to download the spreadsheet, and the CDC pneumonia deaths and pneumonia with COVID deaths are listed here: https://www.cdc.gov/nchs/nvss/vsrr/covid19/index.htm It looks like that's where Briggs was pulling his data from, so I hope he updates it because a lot has been added in the past month that I think might shift some of his conclusions, though I have to go back and read more closely when my head eases up a bit.

    Douglas2 - the Colorado thing was ridiculous. In my day job my team is collecting death and impact data on cancer patients and COVID, so I know how tricky it is to balance all the factors. We don't have to report to any government agency so we don't have political pressures, but you always have to spell out clearly what you're doing and what you're reporting if people are going to be looking at it. To be less than clear when you know how much these numbers are going to be scrutinized is ridiculous.

    ReplyDelete
  8. Thank you, Bethany.

    @DOuglas2 - Colorado was exactly what I was thinking of. People on sites we are familiar with immediately went to "See? The totals might be 25% too high in the whole country!" It illustrates something I had not considered before, that whether something is an anecdote or serious data does have some contextual parts. In the context of trying to precisely track what is happening, both in total and comparing places, Colorado is clearly data. But in the context of people yelling on the internet, it is closer to being anecdote, because it will be so rapidly overinterpreted.

    Something new to worry about an account for, I suppose. Sigh.

    ReplyDelete