Mortality with Meep: Location of Death Data for 1999-2020, Entire United States
Lots of at-home non-COVID deaths in 2020
Emma Woodhouse (who is handsome, clever, and rich, with a comfortable home and happy disposition, seemed to united some of the best blessings of existence, and has lived nearly twenty-one years in the world with very little to distress or vex her. . . and is also somebody pseudonymous on twitter whom I follow) has dug through place-of-death data for Illinois:
For what it’s worth, one need not look at place of death to conclude a lot of excess mortality in 2020 was due to non-COVID causes. I’ve found that out looking at top causes of deaths (drug overdoses way up) and when I look by age group a lot of excess mortality of younger adults wasn’t COVID. I will likely be going over those trends for many years, as will many other people.
That said, I do want to dig into place of death, just to get a grip on where people are dying. This is the first post of at least two, because there is a large amount of data and I need not dump it all on you at once. For this one, I will be looking at a very high level, and not digging into age group, sex, or state-level detail.
Data sets: NVSS Provisional COVID-19 Deaths by Place of Death and Age and WONDER database on deaths 1999-2019, both accessed 19 April 2021.
The data sets gives counts of deaths by the following venues: [the names are a little different between databases, but they have the same categories]
Healthcare setting, inpatient
Healthcare setting, outpatient or emergency room
Healthcare setting, dead on arrival
Decedent’s home
Hospice facility
Nursing home/long term care facility
Other [I’m including “place unknown” in this category]
For 2020, we have death counts split out as:
COVID-19 Deaths
Total Deaths
Pneumonia Deaths
Pneumonia and COVID-19 Deaths
Influenza Deaths
Pneumonia, Influenza, or COVID-19 Deaths
We have results by state, age group, and month. The age groups are bigger than I usually use, but that’s how they serve up the data, and I aggregated historical data to match those groups. If you’re annoyed that ages 50-64 are aggregated together (as I am), then take it up with the CDC.
For the historical data, I can get to far more detail than the 2020-2021 info, but for this and the next post, I am going to keep to COVID deaths (only for 2020-2021) and non-COVID deaths.
Total deaths by location of death, U.S., 1999-2020
Let us start out at the highest level, so that we can look for big trends.
Here are the absolute numbers by place, 1999-2020, stacked column:
Observations:
They didn’t start recording hospice deaths until 2003, so that may have been in the “other” category then. It doesn’t much matter for very recent trends.
You can really see the jump up in total deaths – that’s an 18% increase from 2019 to 2020 in total
With a stacked column graph, you can’t easily see the trends of the various components. So let’s transform to a different type of graph:
Observations:
All places of death saw an uptick in number of deaths in 2020 EXCEPT for hospice.
The 2019-2020 change:
Decedent’s home: +24%
Medical facility, inpatient: +26%
Nursing home: +9%
Other: +24%
Medical facility, outpatient/ER: +13%
Hospice: -11%
Medical facility, DOA: +3%
I note there were various “crossovers” in 2014-2015, when certain places of death increased beyond neighbors
Hospice deaths steadily grew from 2004 to 2015, exceeding Other & Medical facility: outpatient/ER in 2014, but seemed to have plateaued since then.
Deaths at home have been increasing a lot, taking off around 2010. This trend really concerns me.
Note that inpatient hospital deaths had been falling (can be good or bad — discharge ill patients to home, and then they die at home?) and seemed to have flattened out in the period where at-home deaths were taking off. I’m really not liking this trend.
I would not have guessed nursing home/LTC deaths had been flat for a long time. Interesting.
Medical facility: DOA has been pretty low for a long time, and has been decreasing. That may not be good, really, if the reason there aren’t so many DOAs is that the person has been dead for a while when the body was found at their home. Yikes.
It’s good to have a long-term view of how these trends have gone, so you can see just how anomalous 2020 was.
Place of COVID deaths vs. place of non-COVID deaths
Given I have an 18% increase from 2019 to 2020, and that we know COVID was a large piece of that increase (though not 100% of that increase), let’s do a panel of graphs by the place of death, with a breakout of the COVID (vs non-COVID) deaths in 2020.
To make the graphs easier to compare, I’m pushing back death counts only to 2011, so there are 10 full years of data. Also, just because 6 groups nicely and 7 doesn’t, I’m excluding the DOA graph — that’s the smallest of all the groupings, so let us not worry about that one. (Also, the pattern isn’t very interesting).
Here we go:
There are many patterns here I find concerning, but nothing that tells me that under- or over-counting of COVID deaths occurred.
For a few locations, it’s pretty clear that COVID explains almost all their excess deaths: inpatient healthcare facilities and nursing homes. Indeed, it looks like over 100% of the nursing home excess mortality came from COVID, which accords with what I see with excess mortality for older people.
However, there is a lot of excess mortality for people who died at home, and most of that is currently unexplained by COVID.
I don’t think it will be — I think we will find those excess diabetes, heart attack, and ‘unintentional injury’ deaths will have been at home, and because of lockdowns there weren’t other people around to get these people to treatment before they died. This accords with what Emma Woodhouse saw for Illinois – that pattern holds for the entire U.S., it seems.
Then there are the people who normally would have gone to hospice — were they unable to go because of official policy, and thus they died at home? (note: cancer deaths were flat between 2019 and 2020)
In my next post, I will break these data down further by age group so we can see another dimension of the differences.
Underlying spreadsheet (which does have the graph of the DOA deaths)