U. of Chicago COVID-19 interactive data visualization tool

The U. of Chicago has taken my infection rate metric (confirmed cases per million population) to the next level: interactive county-by-county visualization. This shows hot spots that state level data miss. Hot spots are counties with high infection rate that are surrounded by counties with elevated infection rates (this filters outliers, isolated counties with a high infection rate). The U. of Chicago is using the same data source that I have been using in my personal data analysis (1point3acres.com).

https://news.uchicago.edu/story/stat...ncluding-south

The tool allows you to drill down to county level data that includes:
1. Confirmed case count.
2. COVID-19 deaths.
3. Licensed hospital beds
4. Daily new data (cases, deaths, infection rate, death rate)

https://geodacenter.github.io/covid/map.html

The country-wide view can select from 10 different metrics:
1. Confirmed count
2. Confirmed count per 10k population
3. Confirmed count per licensed bed (this is well above 1 for the NYC area)
4. Death count
5. Death count per 10k population
6. Death count per Confirmed count
7-10. Daily metrics

All of this data is available by date since the start of the crisis. You can also compare state-only data to country data to see the dramatic difference between the two.

Very nice!Â* Thanks for sharing.

On 4/6/2020 12:07 PM, 2G wrote:
The U. of Chicago has taken my infection rate metric (confirmed cases per million population) to the next level: interactive county-by-county visualization. This shows hot spots that state level data miss. Hot spots are counties with high infection rate that are surrounded by counties with elevated infection rates (this filters outliers, isolated counties with a high infection rate). The U. of Chicago is using the same data source that I have been using in my personal data analysis (1point3acres.com).

https://news.uchicago.edu/story/stat...ncluding-south

The tool allows you to drill down to county level data that includes:
1. Confirmed case count.
2. COVID-19 deaths.
3. Licensed hospital beds
4. Daily new data (cases, deaths, infection rate, death rate)

https://geodacenter.github.io/covid/map.html

The country-wide view can select from 10 different metrics:
1. Confirmed count
2. Confirmed count per 10k population
3. Confirmed count per licensed bed (this is well above 1 for the NYC area)
4. Death count
5. Death count per 10k population
6. Death count per Confirmed count
7-10. Daily metrics

All of this data is available by date since the start of the crisis. You can also compare state-only data to country data to see the dramatic difference between the two.

--
Dan, 5J

Nice site.

I've been looking at a lot of data. The problem with the confirmed case counts is that they seem to be substantially gated by availability of tests. The rates of confirmed cases by age bracket vary by 40x (older people get a lot more tests, because they get sicker - but I doubt millennia's are 1/40th as likely to get infected - they just don't get symptomatic - or sick enough to justify a test).

So, rates of infection based on under-testing alone are possibly 5x what's reported overall. Then there is the time lag from infection to test result which has averaged around 11 days (7 days to become symptomatic enough to go to the doctor and 4 days to get a test result and biome a confirmed case) , so the reported rates of infection are whatever the daily growth rate is to the 11th power. That can be up to 25x in the rapid growth phase, but probably more like 2-3x now. Once everything peaks the time lag effects are more manageable, but still the mortality rate is a highly confounded metric because the denominator is so uncertain.

Once we get broad, randomly distributed antibody tests we will know a lot better what's going on. In the mean time take a grain of salt on the reliability of the data - depending on which data you are looking at.

Andy Blackburn
9B


On Monday, April 6, 2020 at 11:07:16 AM UTC-7, 2G wrote:
The U. of Chicago has taken my infection rate metric (confirmed cases per million population) to the next level: interactive county-by-county visualization. This shows hot spots that state level data miss. Hot spots are counties with high infection rate that are surrounded by counties with elevated infection rates (this filters outliers, isolated counties with a high infection rate). The U. of Chicago is using the same data source that I have been using in my personal data analysis (1point3acres.com).

https://news.uchicago.edu/story/stat...ncluding-south

The tool allows you to drill down to county level data that includes:
1. Confirmed case count.
2. COVID-19 deaths.
3. Licensed hospital beds
4. Daily new data (cases, deaths, infection rate, death rate)

https://geodacenter.github.io/covid/map.html

The country-wide view can select from 10 different metrics:
1. Confirmed count
2. Confirmed count per 10k population
3. Confirmed count per licensed bed (this is well above 1 for the NYC area)
4. Death count
5. Death count per 10k population
6. Death count per Confirmed count
7-10. Daily metrics

All of this data is available by date since the start of the crisis. You can also compare state-only data to country data to see the dramatic difference between the two.

On Monday, April 6, 2020 at 6:45:59 PM UTC-7, Andy Blackburn wrote:
Nice site.

I've been looking at a lot of data. The problem with the confirmed case counts is that they seem to be substantially gated by availability of tests. The rates of confirmed cases by age bracket vary by 40x (older people get a lot more tests, because they get sicker - but I doubt millennia's are 1/40th as likely to get infected - they just don't get symptomatic - or sick enough to justify a test).

So, rates of infection based on under-testing alone are possibly 5x what's reported overall. Then there is the time lag from infection to test result which has averaged around 11 days (7 days to become symptomatic enough to go to the doctor and 4 days to get a test result and biome a confirmed case) , so the reported rates of infection are whatever the daily growth rate is to the 11th power. That can be up to 25x in the rapid growth phase, but probably more like 2-3x now. Once everything peaks the time lag effects are more manageable, but still the mortality rate is a highly confounded metric because the denominator is so uncertain.

Once we get broad, randomly distributed antibody tests we will know a lot better what's going on. In the mean time take a grain of salt on the reliability of the data - depending on which data you are looking at.

Andy Blackburn
9B


On Monday, April 6, 2020 at 11:07:16 AM UTC-7, 2G wrote:
The U. of Chicago has taken my infection rate metric (confirmed cases per million population) to the next level: interactive county-by-county visualization. This shows hot spots that state level data miss. Hot spots are counties with high infection rate that are surrounded by counties with elevated infection rates (this filters outliers, isolated counties with a high infection rate). The U. of Chicago is using the same data source that I have been using in my personal data analysis (1point3acres.com).

https://news.uchicago.edu/story/stat...ncluding-south

The tool allows you to drill down to county level data that includes:
1. Confirmed case count.
2. COVID-19 deaths.
3. Licensed hospital beds
4. Daily new data (cases, deaths, infection rate, death rate)

https://geodacenter.github.io/covid/map.html

The country-wide view can select from 10 different metrics:
1. Confirmed count
2. Confirmed count per 10k population
3. Confirmed count per licensed bed (this is well above 1 for the NYC area)
4. Death count
5. Death count per 10k population
6. Death count per Confirmed count
7-10. Daily metrics

All of this data is available by date since the start of the crisis. You can also compare state-only data to country data to see the dramatic difference between the two.

Agreed that there is a major limitation on the confirmed cases data, but it is all we have to work with. I expected a quantum jump in this as testing became more available, but that didn't happen, just a very smooth exponential increase. This must be because "confirmed" must include doctor's diagnosis as well as positive test results. On the other hand, deaths are deaths, so you can rely on that data.

Tom

Yup.

Only the mortality data offers reliable trending and it's lagged 2-3 weeks.

I think one of the reasons why we may be seeing the peak in cases a bit later than the initial models predicted is we are seeing a fairly smooth increase in testing availability - adding a bit of a false growth rate. At some point we will be able to test everyone who presents for medical attention - which ought to be a fairly steady proportion of overall cases and so okay as a (still lagged) view of new case trending.

As the testing lag/backlog closes we should have a reasonably decent sense of trending. It won't be until we get broad and randomized testing (antibody tests to look at all infections since the beginning and RNA tests to get rapid identification of current infections for contract tracing and rapid isolation). That we will have any detailed sense of what is really going on - or an ability to contain new outbreaks - which will likely be many.

Life is unlikely to return to normal until sometime in 2021 with (hopefully) a broadly available vaccine. Until then we will all be sitting in a tinder box with a bunch of lit candles and a handful of fly swatters.

Until then we play with Excel spreadsheets and hope.

Andy

On Monday, April 6, 2020 at 9:26:47 PM UTC-7, 2G wrote:
Agreed that there is a major limitation on the confirmed cases data, but it is all we have to work with. I expected a quantum jump in this as testing became more available, but that didn't happen, just a very smooth exponential increase. This must be because "confirmed" must include doctor's diagnosis as well as positive test results. On the other hand, deaths are deaths, so you can rely on that data.

Tom

Hope is not a strategy

Only the mortality data offers reliable trending and it's lagged 2-3 weeks.


Excel's daily ratio between new and active cases seems useful as well. It has been going down from 35% to 10% as the testing ratios are going up. That seems a reliable indicator that the transmission rate is going down.

In other words, the tinder box with only the candles lit is looking good and trending possible.

If behavioral changes to lower R are more than just a fly swatter, then perhaps minimal risks to R like flying solo later in the summer or Fall.

We should about break even this year, pneumonia deaths have fallen to almost zero this winter.

On Tuesday, April 7, 2020 at 12:05:53 AM UTC-7, Andy Blackburn wrote:
Yup.

Only the mortality data offers reliable trending and it's lagged 2-3 weeks.

I think one of the reasons why we may be seeing the peak in cases a bit later than the initial models predicted is we are seeing a fairly smooth increase in testing availability - adding a bit of a false growth rate. At some point we will be able to test everyone who presents for medical attention - which ought to be a fairly steady proportion of overall cases and so okay as a (still lagged) view of new case trending.

As the testing lag/backlog closes we should have a reasonably decent sense of trending. It won't be until we get broad and randomized testing (antibody tests to look at all infections since the beginning and RNA tests to get rapid identification of current infections for contract tracing and rapid isolation). That we will have any detailed sense of what is really going on - or an ability to contain new outbreaks - which will likely be many.

Life is unlikely to return to normal until sometime in 2021 with (hopefully) a broadly available vaccine. Until then we will all be sitting in a tinder box with a bunch of lit candles and a handful of fly swatters.

Until then we play with Excel spreadsheets and hope.

Andy

On Monday, April 6, 2020 at 9:26:47 PM UTC-7, 2G wrote:
Agreed that there is a major limitation on the confirmed cases data, but it is all we have to work with. I expected a quantum jump in this as testing became more available, but that didn't happen, just a very smooth exponential increase. This must be because "confirmed" must include doctor's diagnosis as well as positive test results. On the other hand, deaths are deaths, so you can rely on that data.

Tom

One immediate problem is the accuracy of the current testing. One our club members has CV. Multiple doctors giving him medical care say he had all the symptoms. He was extremely sick in ICU (and feeling better now). However, 2 tests came back negative. After talking to his health care providers about the false negatives, their response was it appears to be at a rate of about 20-25%. As you say, better and broader testing down the road is needed, but the current numbers are "in the books".

On Tuesday, April 7, 2020 at 9:18:38 AM UTC-5, wrote:
We should about break even this year, pneumonia deaths have fallen to almost zero this winter.

gregg..., Where did you get your information from, or are you just messing with us? What do you think people are dying from during these influenza and SARS-related pandemics? Victims are put on ventilators because of....? Really, give us a few citations to support your claim please. Here's one from me.

https://www.rochesterregional.org/ne...lu-season-2020