What are the challenges in accurately reporting and interpreting COVID-19 data?

[jim mcnamara]

TL;DR Summary

COVID data comments

�

Source	Deaths	Data timestamp
https://www]cdc.gov/coronavirus/2019-ncov/cases-updates/cases-in-us.html	175,651	Aug 23 2020 12:15PM EDT
https://www]worldometers.info/coronavirus/#countries	180,724	August 24, 2020, 16:47 GMT
Johns Hopkins U ARCGIS*	176,901	8/24/2020, 10:27:56 AM
https://covidtracking.com (Atlantic magazine)	168,858	Aug 24 at 12:19 pm ET
* Dynamic URL, use Google

This kind of data mishmash is common in medical reporting, because local governments control how, when, and where vital statistics are recorded.

And for the US, some political input on evaluating death certficiates seems possible. In this example, the reporting entities are compiling data from lots of sources and it can get confusing for them and us.
For the US this might be 48 states plus Washington DC. Or not.

Example:
Local government has a holiday, FooBar Day on a Monday, and so they do not report any deaths. Coroners office closed. The next business day, the missed deaths are reported and the data compilers do not necessarily know about holidays. So Tuesday after the Monday looks anomalous.

Point:
You may want to avoid getting 'wrapped around the axle' when you see things like this. AFAIK all of the entities are trying to follow their own protocols. Often you may need to take a look at something else to see if the data is okay.
Example excess deaths:
Normally on the first Monday of July, FooBar County reports an average 20 all cause deaths, with a sigma of 3. But you have 38 deaths all from some catchall category. You might want to assume there are excess deaths: $38 - ( 20 + \sigma*2) = 12$ excess deaths. (yes, 2 $\sigma$ not 3, we used this as a a guideline for a vital statistics project for past epidemics)

There are a lot of other things going on. The above is just a simplification. I use the CDC site

 

[Vanadium 50]

This doesn't look too bad to me. If you remove the outlier, the other three have a standard deviation equivalent to 2 days. That seems pretty good. Especially compared to other causes of death: I challenge you to find an official suicide number that's newer than a year old. Maybe two. (I have no idea why the suicide lag is as long as it is)

 

[jim mcnamara]

Re: suicides. The part of the lag I understand is what some places call an adjudication step. In my county all suicides are handled by the Coroner's office and are treated as a potential criminal act. Sometimes the case gets referred back to the County Police if it looks criminal.

Suicides on local reservations is when the FBI gets involved from the get-go. Tribes do not tolerate autopsies, so it gets really sticky. Our local FBI guys have to let autopsies slide most of the time. Then it becomes a paper trail. If the Native American corpse is found off the Res, it is a local police matter. But generally still no autopsies.

After all this, if the case clears all the hurdles, then a judge has to rule it as an official suicide. Then, at last, it can become an official statistic.

Interesting story: the reservation I lived on had a tradition. 'Sleep with the bears' is a crummy translation.

Very occasionally -- Older folks, usually men, went out at sunset in extreme cold, climbed to the top of a mesa, took off all clothing, and sat down. Forever.

 

[Vanadium 50]

Re: suicides. The part of the lag I understand is what some places call an adjudication step.

The AP just published an article that military suicides are up 20%. The military knows, loves, and understands paperwork, so they are apparently able to process this more quickly. Extending this increase to the population at large means an additional 10,000 in the US. This would appear as "Covid excess deaths".

 

[jim mcnamara]

These two statements from the CDC site disagree with you. non-Covid, yes. Excess deaths get lumped using DC data.

Standard DC used by US Armed Forces:
https://www.cdc.gov/nchs/data/dvs/death11-03final-acc.pdf

Look at item #37. Specifically notes suicide deaths.

The Armed Forces DC numbers reported to the CDC would inflate excess deaths but would NOT affect Covid-19 fatality reporting. Correct. See why below.

Excess deaths should be larger than the Covid-19 reported deaths on the CDC site. And they are.
So the CDC this to get around the problem:

Comparing these two sets of estimates — excess deaths with and without COVID-19 — can provide insight about how many excess deaths are identified as due to COVID-19 ...

Read more here:
https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm

No misplaced 10K deaths as far as the CDC believes. What those solders actually did die from is always based physicians opinions and sometimes autopsies. So who knows perfectly...

@OmCheeto the current page link directly above has more links to datasets. Don't know if they are useful. FWIW the standard US DC in the top link could potentially be of use.
These vary - a lot - even between counties. And suicide is a mess -- sometimes it takes more than a year to substantiate the claim.

 

[berkeman]

Tribes do not tolerate autopsies, so it gets really sticky

I didn't know that.

 

[Vanadium 50]

I don't think I am saying anything they aren't saying (other than numbers from the AP. They say "Upward trends in other causes of death (e.g., suicide, drug overdose, heart disease) may contribute to excess deaths in some jurisdictions." and "These deaths could represent misclassified COVID-19 deaths, or potentially could be indirectly related to COVID-19."