COVID Quickies: The New York Times, Old People Dying, WONDER, and Fixing Graphs
Seriously, NYT, what is up with your graphs
Whew, a lot to try to push out before the 12 days of Christmas (and do I have plans for Christmas! (do I? (YES!))
First, let me lead with the good news: THE 2020 DATA ARE IN!
Let all the actuaries rejoice!
Okay, we’ve had 2020 data for some time, but these are the finalized CDC data.
Here’s a video:
Yes, I will later get into the “Is it COVID or isn’t it?” …. but after the Christmas season. I have other plans for the 12 Days of Christmas….
Fixing a stupid graph at the New York Times
If you want to see how I got drawn into this, you can read this Twitter thread and then I read the NYT article in question and then did this thread. It’s annoying trying to embed tweets into my own posts, so let me show you the article and then the graph… and I’m going to take it apart and making it better.
The article in question
NYT: As U.S. Nears 800,000 Virus Deaths, 1 of Every 100 Older Americans Has Perished
As the coronavirus pandemic approaches the end of a second year, the United States stands on the cusp of surpassing 800,000 deaths from the virus, and no group has suffered more than older Americans. All along, older people have been known to be more vulnerable, but the scale of loss is only now coming into full view.
Seventy-five percent of people who have died of the virus in the United States — or about 600,000 of the nearly 800,000 who have perished so far — have been 65 or older. One in 100 older Americans has died from the virus. For people younger than 65, that ratio is closer to 1 in 1,400.
…..
After the first known coronavirus death in the United States in February 2020, the virus’s death toll in this country reached 100,000 people in only three months. The pace of deaths slowed throughout summer 2020, then quickened throughout the fall and winter, and then slowed again this spring and summer.
Throughout the summer, most people dying from the virus were concentrated in the South. But the most recent 100,000 deaths — beginning in early October — have spread out across the nation, in a broad belt across the middle of the country from Pennsylvania to Texas, the Mountain West and Michigan.
These most recent 100,000 deaths, too, have all occurred in less than 11 weeks, a sign that the pace of deaths is moving more quickly once again — faster than at any time other than last winter’s surge.
By now, Covid-19 has become the third leading cause of death among Americans 65 and older, after heart disease and cancer. It is responsible for about 13 percent of all deaths in that age group since the beginning of 2020, more than diabetes, accidents, Alzheimer’s disease or dementia.
Okay.
I am trying to be as charitable as possible, given the TIME OF CHRISTMAS CHEER, to actually do the work they were trying to do with this article and the graph I’m about to yell about.
I believe this is the point they’re trying to make in the article and the graph below: yes, COVID is more harsh on old people (as is life in general). But even after the vaccines, and old people being more vaccinated than young people, they’re still bearing the brunt of COVID deaths!
Yes, yes they are. Thank you for noticing. This is actually an important point, which should be pointed out during every flu season, too.
And no, I’m not trying to minimize the COVID impact nor making a joke about flu and pneumonia deaths. I don’t joke about flu and pneumonia deaths. (I wrote that post in 2018, btw.)
The graph in question
This is the graph accompanying the article.
In my initial reaction to this graph, I thought they were talking about all deaths, but this graph isn’t going from 0% to 100% of deaths, but 0% to 100% of COVID deaths. That should be a little obvious due to the extreme wave structure of this graph.
First fix needed: better labeling of the vertical axis
Then there are a few items superimposed on this graph, in a “helpful” way.
Dates are put along the horizontal axis, which is actually helpful. Thank you. The horizontal axis is spaced out appropriately for time (that is, no weird scaling — 1 day = the same amount of horizontal length)
These dates specifically marked on the graph, with accompanying vertical lines, mark each 100K of COVID deaths have been “achieved”. Hmmm. That makes this graph more difficult for people to read, because people may not realize what’s happening between May 27, 2020 and Sept 22, 2020 (i.e., slowdown in COVID deaths) versus, Dec 14, 2020 to Jan 19, 2020 and Jan 19, 2021 to Feb 22, 2021 (when we had our worst COVID death wave thus far).
What might have helped would have been two graphs, but I guess they didn’t have the space…. no. That’s bullshit. They went with a confusing graph over a graph I’m about to do. Why they did it this way, hell if I know.
Finally, they thought it was meaningful to put a dividing line between under age 65 and over age 65 for deaths, and then have a horizontal line across to show the average percentage over the pandemic. Because that’s super-meaningful.
(Note: while distinguishing between “old” and “not old” deaths is useful, and you have to pick an age (65 is fine), doing that “average” line across the pandemic was not useful.)
Oh, and they mark the “past 700K COVID deaths” as if that’s important in any way.
(Note: It’s not)
I’m going to fix this graph in multiple ways, and we’ll see if we can make the point I THINK they’re trying to make.
And then show why it doesn’t show what they think.
Let’s catch those waves again!
Let’s start with our base data, which are COVID deaths by age grouping. I will not yet do a under-65 and over-65 split. I think there are interesting trends here we shouldn’t yet flatten. You may remember this from a post a few months back, in which I was arguing that the vaccines really were cutting risk. I had good data through September 2021 as of October 18, and I think the data should be looking pretty good through the end of November now, if we focus solely on COVID deaths.
Here is the source of data: Provisional COVID-19 Deaths by Sex and Age – last updated 22 Dec 2021. They do have some December 2021 deaths, and if you promise to realize there will be additional COVID deaths yet to come, I will show the two full years of death data by the age groups.
First, graphing the deaths, which I will graph monthly, because I just think it’s easier to look at.
So we see that fourth wave was much worse than the summer 2020 wave.
Now let’s change to percentages as with the NYT graph:
So yes, it looks like that after that summer wave crested back September, the percentage of COVID deaths that were for age 65 and over, which had been receding, started increasing again.
But because the NYT just went with under age 65/over age 65, you didn’t really get to see where the movement was.
Looking at the real trend of interest
Let me do one more graph, switching these percentages to lines, and maybe we can see what’s going on:
FIXING MY OWN GRAPH (I forgot to label the 40-49 line, and the 30-39 label got misaligned):
Iiiiiiinteresting.
Note that the age 85 and over, which used to be the predominant group of COVID deaths, receded, and yeah, it’s coming back up, it seems a lot of the movement is coming from the age 50-64 bin, which is the older GenX/younger Boomers bin.
Having seen the lag in vaccination rate for that age group, it may be that they’re just finally catching up with the seniors on vaccination rates, and I see that that group just recently reached 90% getting at least one vax dose. So that may be more related to the old folks representing a higher percentage of deaths more than anything else.
When you’re seeing that one group is an increasing percentage of “bad stuff” happening to them in the population overall, you might want to check if it’s just relative in that other groups are simply improving more rapidly. The seniors had their rapid improvement earlier as they had first access to vaccines and boosters.
Don’t be surprised that once the middle-aged folks caught up in vax rates that they’d be back in a similar ratio… though they’re not, actually. The 50-64 group is still disproportionately high in their share, compared to earlier in the pandemic.
By squishing all the “under age 65” together as if that were meaningful (nope) and all the “over age 65” together as if that were meaningful (nope) – they lost a lot of useful information. And then they added a bunch of detail re: when certain round hundreds of thousands of COVID deaths occurred, and specific dates, which wasn’t useful information. Also, left out useful identifying info in labeling their graphs, and put a bunch of junk text around it.
So, lots of problems with the NYT graph.
I don’t even know what point they’re trying to make other than, even after vaccination, there’s still a huge age differential in mortality risk due to COVID.
Which isn’t that unusual if you compared it to similar diseases.
Why it took such a confusing graph to make such a simple point, don’t ask me.
Working with WONDER: Age groups
This is a little old — it doesn’t have the 2020 data.
But if you go to WONDER right now, you can grab the 2020 data yourself! Try it out!
The spreadsheet I produced in the video can be found here: Dropbox link to view and download the spreadsheet.
Watch this space during the 12 Days of Christmas — more goodies to come, I promise!