Excess Mortality Redux: Investigating the Underestimation of COVID-19 Deaths

[jim mcnamara]

TL;DR

Large dataset and supporting information on under reporting Covid deaths. The results for the US, for example, add about 400k deaths to the total

This is interesting from a method point of view and as a source of mortality data sets with multiple links, sources. Plus explanation.

The underlying message seems to be, IMO, under reporting mortality to varying degrees in a pandemic situation seems to be inevitable. The magnitude of deaths is estimated. Estimates are just that, estimates.

http://www.healthdata.org/special-a...covid-19-and-scalars-reported-covid-19-deaths

For example, the IMHE attributes an additional 400k deaths to Covid in the US, as derived from excess mortality figures. After seeing the problems with mortality reporting in the current (as of 01-07-2021) surge in India, under reporting can be seen. The magnitude is the question.

The concept, simplified:

1. for the past 10 years, say 2009-2019, we know that 2 million people died each year
    with year-to-year ±0.05 million differences.
2. in 2020 2.25 million died. 
3. Excess deaths for 2020 is (2.25 million  minus 2.05 million), which gives 0.20 million.
4. 0.10 million death certificates list Covid, another 0.10 million do not list Covid, 
    but things like  'natural causes'
5. So, what do we do with the leftover 0.10 million deaths due to something else?

Answer from IHME is some of those leftover deaths are related to Covid.

For a different approach to this kind of problem try to get the correct number deaths due to hypothermia, during the the extreme February 2021 cold snap in Texas. Good luck

See what you think.

@OmCheeto may find the IHME data sets interesting.

 

[Vanadium 50]

I absolutely agree that excess deaths tell us something. I disagree with the people who believe that they are somehow the "gold standard" of estimation.

As the authors of the paper point out, some excess deaths are not attributable to the virus at all. Murders in Chicago (and gunshot injuries) are up by 50%, nationally the numbers seem not to be final but it's above 25%. Those are excess deaths but not Covid deaths.

In the other direction - a friend's husband died of Covid while he had advanced pancreatic cancer. He was unlikely to make it through the year. That's a Covid death that's not an excess death.

I think the real value of the sample is not in counting totals, but at looking at distributions to identify places that need as closer look. Does the excess death population match the Covid death population everywhere? If not, what does that tell us?

 

[Rive]

Also a problem with the 'raw' excess death kind of statistics is, that the base value may change due preventive measures in effect.

Just the most obvious example: the flu were gone for this season. Well, at least where face mask was worn and distance were kept. So with flu deaths out of the picture, what is the 'excess' we can account for covid? Compared to what base?

Next, home office and traffic accidents... And so on.
Last year we (mid-europe) had ~ 20000 Covid victims, but on yearly average it came out around the limits of usual fluctuation: mostly accounted on the lower base.