Read on Twitter
A 
on COVID-19 deaths
1. trends in COVID-19 deaths in FL
2. trends in "excess deaths" using CDC estimates
In summer "surge" in FL:
- 40 straight days w/ 7-day avg deaths >150
- 20 straight days w/ 7-day avg deaths >200
- peak was 240 (8/4)
- peak 7d avg was 227 (8/5)
1/
on COVID-19 deaths 1. trends in COVID-19 deaths in FL
2. trends in "excess deaths" using CDC estimates
In summer "surge" in FL:
- 40 straight days w/ 7-day avg deaths >150
- 20 straight days w/ 7-day avg deaths >200
- peak was 240 (8/4)
- peak 7d avg was 227 (8/5)
1/
In Jan in FL:
- now at 20 straight days w/ 7-day avg deaths >150 (1/3 to 1/26)
- peak at 204 (1/15)
- peak 7d avg was 187 (1/18)
Reporting still coming in, so these numbers likely to rise. But likely not to match summer peak.
Why?
2/
- now at 20 straight days w/ 7-day avg deaths >150 (1/3 to 1/26)
- peak at 204 (1/15)
- peak 7d avg was 187 (1/18)
Reporting still coming in, so these numbers likely to rise. But likely not to match summer peak.
Why?
2/
Well, the "peak" of "current hospitalizations" in mid-Jan did not get as high as the summer surge.
And as you've likely heard me say, 7d average hospitalizations are very predictive of 7d average deaths 5-7 days later.
3/
And as you've likely heard me say, 7d average hospitalizations are very predictive of 7d average deaths 5-7 days later.
3/
In fact, the [poor] running joke of just wait 2 weeks (for deaths) to accumulate plays out.
Look at the shape of the current hospitalizations and actual deaths from 7/11 to 1/16 (when hosp trends start to decline).
The deaths follow that hosp curve to a tight fit.
4/
![In fact, the [poor] running joke of just wait 2 weeks (for deaths) to accumulate plays out.Look at the shape of the current hospitalizations and actual deaths from 7/11 to 1/16 (when hosp trends start to decline).The deaths follow that hosp curve to a tight fit.4/](https://pbs.twimg.com/media/EuZQ3qBVoAAFEPK.jpg "In fact, the [poor] running joke of just wait 2 weeks (for deaths) to accumulate plays out.Look at the shape of the current hospitalizations and actual deaths from 7/11 to 1/16 (when hosp trends start to decline).The deaths follow that hosp curve to a tight fit.4/")
Look at the shape of the current hospitalizations and actual deaths from 7/11 to 1/16 (when hosp trends start to decline).
The deaths follow that hosp curve to a tight fit.
4/
In fact, the ratio of deaths to current hospitalizations has remained fairly steady since July, ranging from 0.0201 to 0.0288.
As in if the "factor" is 0.0250, then 6000 current hospitalizations = 150 deaths 5-7 days later.
5/
As in if the "factor" is 0.0250, then 6000 current hospitalizations = 150 deaths 5-7 days later.
5/
In general, lag in death reporting still much improved. Most days have ~70% of deaths reported having occurred within 2 weeks and ~90%+ within 30 days.
6/
6/
Now, CDC's excess deaths.
This is when a comparison is made between average weekly deaths in the years preceding 2020 & deaths occurring since.
Obviously, the pandemic has resulted in "excess deaths" due to all-causes, relative to what we would have expected sans COVID.
7/
This is when a comparison is made between average weekly deaths in the years preceding 2020 & deaths occurring since.
Obviously, the pandemic has resulted in "excess deaths" due to all-causes, relative to what we would have expected sans COVID.
7/
Best lower and upper bound estimate for entire US is between 439,435 and 533,307 excess deaths since February 2, 2020.
But everyone loves to compare states, so...
8/
But everyone loves to compare states, so...
8/
As you've heard me say with other metrics, don't compare VOLUME of excess deaths ALONE.
In general, states with larger populations will have a larger number of excess deaths, as you can left in this side-by-side of excess death estimates (LEFT) and population (RIGHT).
9/
In general, states with larger populations will have a larger number of excess deaths, as you can left in this side-by-side of excess death estimates (LEFT) and population (RIGHT).
9/
So, I like the % excess deaths, which for each state/jurisdiction assessing the number of excess deaths as a % above what would have been expected.
For this measure, top 5 (worst) are NYC (reported separately from the rest of NY), NJ, DC, AZ, and MI.
10/
For this measure, top 5 (worst) are NYC (reported separately from the rest of NY), NJ, DC, AZ, and MI.
10/
Also, the shape of the excess death curve over time will be different depending on when states were impacted. Look below at Florida (top left), NYC (top right), Texas (bottom left), and entire US (bottom right).
Please note different Y axes in 4 figures.
11/
Please note different Y axes in 4 figures.
11/
Lastly,
These estimates are impacted by death reporting lag. So recent estimates of excess deaths will increase.
In fig, each line is a new CDC weekly report.
For 12/9 excess deaths
5238 on 12/23 rpt
15203 on 1/6 rpt
20242 on 1/17 rpt
22351 on 2/10 rpt
12/
These estimates are impacted by death reporting lag. So recent estimates of excess deaths will increase.
In fig, each line is a new CDC weekly report.
For 12/9 excess deaths
5238 on 12/23 rpt
15203 on 1/6 rpt
20242 on 1/17 rpt
22351 on 2/10 rpt
12/
If people are interested in excess deaths, I have a bunch of visualizations I could add to my site. We'll see what the interest is. Thanks for humoring me for another !!
13/end
!!13/end
