WANT TO REALLY TRACK RECENT TRENDS?
- for your county?
- for meaningful age groups?

Seen the @nytimes display of recent trends 👇 but wondered about heterogeneity across age groups that may be differentially impacted by decisions we make?

Now you've got a viz for that.

🧵1/
First viz 👇 (#6 on my CASES page) looks bland enough, but contains good info.

The default is statewide.

Age groups (9) make up the rows.

"Case date" makes up the columns, from most-to-least recent.

The # in each cell is the TWO-WEEK CHANGE IN DAILY CASES.

The what?

2/
The calculation:

I want to know 2-wk trends from 8/24 to 9/7.

On 8/24, we calculate the 7-day rolling avg daily cases (to remove daily variability)

On 9/7, do same thing.

What's in the cell for 9/7 is the diff in those numbers, expressed as a % change.

Tooltip helps 👇

3/
For example, in the 18-24 population statewide, as of Aug 24, there had been a 42% decrease in avg daily cases (+ tests) in the previous 2 wks.

Fast forward 2 wks to 9/7, and there was a 57% increase in avg daily cases relative to 8/24.

From a period of ⬇️to a period of ⬆️

4/
How does this look during a period of global increase, as was the case in June?

Just scroll to the right.

The more 🔴 the higher the % increase in prev 2 weeks.

Growth started with younger age groups, and then made its way to more pronounced % increase in older groups.

5/
But that's statewide - we always say this is not a one-size-fits-all situation.

Scroll to the viz below (#7).

The default has counties making up the rows, based on the experience of all age groups combined.

Just a look at the first flew counties shows the heterogeneity.

6/
Then you can filter by age.

AND, since daily cases estimates can have high temporal variability in small counties, especially when filtering by age, I included a filter based on county pop.

Below I filter to the 23/67 counties w/ 250,000+ pop.

Still big heterogenity.

7/
How might one use this?

Let's focus on all age groups except college-aged. By Aug 24, all large counties had been experiencing ⬇️ in daily cases in previous 2 weeks (all 🟢)

Then, during September, a few ⬆️ here and there, but nothing pronounced.

8/
What if we look only at college-aged?

By Aug 24, all large counties had been experiencing ⬇️ in daily cases in previous 2 weeks (all 🟢)

BUT, look at September. 🔴🔴🔴🔴🔴🔴🔴 for a higher proportion of counties...but also not in all of them.

9/
In Alachua county, the hover-based tooltip for the 18-24 group is 👇

The number in the cell for 9/14 is 1926.

A 1926% increase in the 7-day avg daily cases on 9/14 compared to 8/31.

From 5 daily cases to 98 daily cases. Hence the deep 🔴 color.

10/
Could do the same with vulnerable populations, at least based on age (e.g., 80+)

This will be important to monitor when community spread ⬆️ in younger groups, to see if it "reaches" vulnerable groups (hopefully not).

👇is a current viz for 80+ in 250k+ counties.

11/
A deeper dive MUST OCCUR.

For ex, previous fig might have concerned you about Volusia county.

But the tooltip tells you the cases just went from ~3 to ~5. A big % increase, but not absolute.

I could then look at an epidemic curve in the 80+ group to confirm the small #

12/
So, I hope this is a nice tool to track 2-wk changes in daily cases for subgroups of interest.

This is different from the @nytimes in that:
- they use the date cases are reported, I use the date of + test
- they provide state & county granularity, I include age specificity

13/
As we await a statewide school-based report, and try to leverage district or county-specific dashboards being created...

...this should facilitate easier visual inspection of changing trends in age groups and regions of interest to us all.

14/
You can follow @JasonSalemi.
Tip: mention @twtextapp on a Twitter thread with the keyword “unroll” to get a link to it.

Latest Threads Unrolled: