Mining CDC data

[toro1]

I was looking for some particular COVID date and found some interesting data on the CDC web site here-

https://www.cdc.gov/nchs/nvss/vsrr/covid19/index.htm

There are 2 tables that show total expected US deaths from 2/1/20 to 5/16/20 as well as showing deaths attributed to COVID. The first table shows the weekly summary for the entire nation and the second table breaks down the state totals.

I found the following things interesting-

In the First Table, the CDC shows the total US deaths for the above time frame and then compares it to the expected death, based on an average of the last 2 years during the same time frames.

In 2020 there were 907,069 deaths from all causes from 2/1-5/16 of which includes 69,788 deaths from COVID. In comparing that to the average from the last 2 years, there are 1% more deaths in 2020 than what would be expected. That is 9070 more deaths than expected even though COVID “caused” an extra ~70K deaths.

If you look at the last 2 weeks shown, the death totals are significantly lower than what was expected with the last week coming in at 30% of what was expected. I can venture some guesses to the cause of this, (COVID sped up some deaths?) but it is speculation.

In looking at the State Data table, 33 states are running below expected death totals including California, whose deaths from all causes is running at 98% of what is expected.

However, if you look at NY, NYC, NJ, Mass, and Ill, their deaths are all well above what was expected with NYC leading the way at 234% of expected.

I’m still mining the data, but comments from the people smarter than I are welcome. Based on the CDC data, my initial thoughts are, outside of the NY/NJ/Mass corridor even the CDC data does not support the hysteria.

[ACfixer]

However, if you look at NY, NYC, NJ, Mass, and Ill, there deaths are all well above what was expected with NYC leading the way at 234% of expected.

I'm guessing the sending CV19 patients to nursing homes wasn't the brightest move.

[smashycrashy]

If you look at the last 2 weeks shown, the death totals are significantly lower than what was expected with the last week coming in at 30% of what was expected. I can venture some guesses to the cause of this, (COVID sped up some deaths?) but it is speculation.

Technical note explains why.. the last 2 weeks are significantly different:
*Data during this period are incomplete because of the lag in time between when the death occurred and when the death certificate is completed, submitted to NCHS and processed for reporting purposes. This delay can range from 1 week to 8 weeks or more, depending on the jurisdiction, age, and cause of death.

You are comparing a period of time in the past which the lags are through the system so the data is up to data with the most current weeks which are the least up to date parts.

[as_rocketman]

As you mine the data, also keep in mind that the various restrictions and social changes as a result have led to decreases in other causes of death. The flu season ended early. Traffic fatalities and homicides are way down. Industrial accidents are way down.

And, yes, there is also some opposite effect -- additional deaths due to suicide or deferred health care, those are in there too. It's a complex problem.

As for the "hysteria," that's an even more difficult calculation. For this you need to compare what happened vs. the potential alternative. After that, you need to establish what level of response was prudent in your own subjective terms.

We are still losing 1,300 people per day to this thing. In terms of IFR and reproduction number, it outstrips the 1968-70 influenza outbreak by about an order of magnitude -- and that obviously earned a little hysteria, since the deniers still hold it up as a yardstick... SARS-CoV-2 definitely needs to be taken seriously. How to respond to it, though, is largely an individual calculation.

[balgor]

I was looking for some particular COVID date and found some interesting data on the CDC web site here-

https://www.cdc.gov/nchs/nvss/vsrr/covid19/index.htm

There are 2 tables that show total expected US deaths from 2/1/20 to 5/16/20 as well as showing deaths attributed to COVID. The first table shows the weekly summary for the entire nation and the second table breaks down the state totals.

I found the following things interesting-

In the First Table, the CDC shows the total US deaths for the above time frame and then compares it to the expected death, based on an average of the last 2 years during the same time frames.

In 2020 there were 907,069 deaths from all causes from 2/1-5/16 of which includes 69,788 deaths from COVID. In comparing that to the average from the last 2 years, there are 1% more deaths in 2020 than what would be expected. That is 9070 more deaths than expected even though COVID “caused” an extra ~70K deaths.

If you look at the last 2 weeks shown, the death totals are significantly lower than what was expected with the last week coming in at 30% of what was expected. I can venture some guesses to the cause of this, (COVID sped up some deaths?) but it is speculation.

In looking at the State Data table, 33 states are running below expected death totals including California, whose deaths from all causes is running at 98% of what is expected.

However, if you look at NY, NYC, NJ, Mass, and Ill, their deaths are all well above what was expected with NYC leading the way at 234% of expected.

I’m still mining the data, but comments from the people smarter than I are welcome. Based on the CDC data, my initial thoughts are, outside of the NY/NJ/Mass corridor even the CDC data does not support the hysteria.

The CDC death data is lagged by up to 2 weeks so the most recent data is incomplete. Considering more than 1/2 the deaths from covid are concentrated in 5 states, it's not surprising those states lead the way in excess deaths and that there are a lot of states below, right at or only slightly above their long term average.

Nytimes did a state by state analysis:

https://www.nytimes.com/interactive/2020/05/05/us/coronavirus-death-toll-us.html

The one thing these excess death charts tell me, is that the excess deaths were caused by covid, not the lockdown. Many states have locked down, had no spike in Coronavirus deaths AND no spike in excess deaths, which tells me the lockdown isn't killing NET (i.e. car/industrial accidents lower, suicides/deferred cared deaths up, but the effects seem to be a wash) people as some here have suggested

[Jedediah Munroe]

I suppose the CDC uses at two year average for comparison because it 2 is the minimum sample while still being able to call it an average? I would bet if we averaged at least a couple decades the current gross death rate would not be an outlier.

[ShockBros]

Terrific! I was just discussing analyzing this same data today and here you are with it!

Can we go back 10 years on the numbers and answer this question?

Thank you Toro you are a true gentleman and a scholar.

[toro1]

To those that say the CDC is lagging, since the comparison is to like weeks from the previous years, it seems the lag would also exist there and so I would expect the comparison to be valid. One thing I do not know is, does the CDC report the figures as the date they received the death certificate or does the data get folded back in to the death date? If its the latter, then those numbers should rise with this weeks update. As for using a longer duration data set, this is the only set I have found so far.

Balgor, Thanks for the link and it looks like they have already done what I was going to do next. I had done NYC already and from the data it appeared for NYC probable COVID deaths were under reported. I have been one of the people that suspected over reporting and so the data seen here is pointing towards under reporting for the NE corridor which is what you have been saying for a while. That being said, there is still the question of who died from COVID and who died with Covid.

ASR, Good points although the next step is to try and get to the basic question of what Covid deaths might be from a group that would have died from other things even without COVID. A detailed look at the last columns needs to be done to understand that aspect.

Thanks Shockbros, but my wife might take exception to your gentleman/scholar comments.

[ACfixer]

It's an interesting way of looking at things OP. How many people are dying from all causes vs. trying to hash out the under/over reporting and also the with/from thing. If the same amount of people die this year in the USA that die every year (within a margin of error of course), well then what does that mean? Well IMO that means people dying with CC19 vs. from, if it's significantly higher, that leads to the from CV19 conclusion. Just the way I see it.

[toro1]

It's an interesting way of looking at things OP. How many people are dying from all causes vs. trying to hash out the under/over reporting and also the with/from thing. If the same amount of people die this year in the USA that die every year (within a margin of error of course), well then what does that mean? Well IMO that means people dying with CC19 vs. from, if it's significantly higher, that leads to the from CV19 conclusion. Just the way I see it.

With all of the competing information inputs it's the only way I could figure out how to get a top level assessment of what is truly going on. I kept thinking of the adage if the facts are on your side, argue the facts, if the facts aren't muddy the water. For most of this the water always seems to be muddied.

For the most part I agree with your last statement but there are a couple of things that ned to be kept in mind while looking at the data. One is what ASR said above about other deaths being lower, but there may also be an effect from the better hygiene and distancing people have been practicing. However, it isn't clear how the stay at home where the virus could be easily passed will work into things. I bring this up because approximately half of the deaths around the country are in care facilities where people are sequestered closely together.

I don't plane to look into that much, but it is something to keep in mind. For those that want to take a look at past data for flu and pneumonia deaths, you can go here-

https://www.cdc.gov/flu/weekly/weeklyarchives2017-2018/Week15.htm

Go to the graph and then click on the "View Chart Data" line under the graph. It shows the actual numbers for the flu and pneumonia deaths and you can compare them to the numbers we are seeing now.

[Jedediah Munroe]

I was wanting to look at the numbers myself about a week and a half ago using only cdc and ca.gov website published numbers and not using most recent couple weeks of data due to poor data (due to usual lag time in reporting deaths). Looks like almost 300,000 people die every year in California.
My rudementary calculations gave the following perspective for 2020 calendar year.
YTD about 70,000 people died from all causes, just under 3,000 had reported died from COVID-19

[smashycrashy]

To those that say the CDC is lagging, since the comparison is to like weeks from the previous years, it seems the lag would also exist there and so I would expect the comparison to be valid. One thing I do not know is, does the CDC report the figures as the date they received the death certificate or does the data get folded back in to the death date? If its the latter, then those numbers should rise with this weeks update. As for using a longer duration data set, this is the only set I have found so far.

The numbers are for not when they received the data, that would make no sense from a historical record perspective. The date is when the person deceased and there is significant lag in the data.

Thus, as I said previously, the data for last year is valid up to date data but any data for the last 8 weeks is susceptible to rise up. The closer the week is the more susceptible to significant upward revision. So you were looking at the last two weeks, the least valid data in the data set. The NVSS is the most lagged of the data set but the most accurate. The other death counts you see are compiled from the individual health departments reporting deaths (as opposed to getting the death certificate into the system like the NVSS data) so it is more timely but has the possibility of small errors and you have a tough time discerning date of deaths and the like for comparison over time periods.