Read on Twitter
A basic (but important) epidemiology 
on comparisons.
In this case, comparing unadjusted COVID-19 death rates across states, counties, or other jurisdictions.
I will illustrated using all 24 counties in FL with an estimated population of 250,000+
1/
on comparisons.In this case, comparing unadjusted COVID-19 death rates across states, counties, or other jurisdictions.
I will illustrated using all 24 counties in FL with an estimated population of 250,000+
1/
By now, most people know that you can't just compare the total number of COVID-19 deaths because counties with bigger populations will have the opportunity for more deaths.
The counts are good for public health burden, but not to compare risk of mortality.
2/
The counts are good for public health burden, but not to compare risk of mortality.
2/
So, we account for differences in county population sizes by calculating a mortality rate, usually expressed as:
# deaths
----------- x 100,000
total pop
If that's the case, then the 5 highest mortality rates are in:
1. Escambia
2. Dade
3. Marion
4. St. Lucie
5. Palm Beach
3/
# deaths
----------- x 100,000
total pop
If that's the case, then the 5 highest mortality rates are in:
1. Escambia
2. Dade
3. Marion
4. St. Lucie
5. Palm Beach
3/
What's the problem?
Well, we know that age is STRONGLY tied to likelihood of death from COVID-19.
37.3% of Sarasota County's pop is 65+
12.3% of Orange County's pop is 65+
A higher proportion of Sarasota's pop is at higher risk of death, so it is not an
to comparison
4/
Well, we know that age is STRONGLY tied to likelihood of death from COVID-19.
37.3% of Sarasota County's pop is 65+
12.3% of Orange County's pop is 65+
A higher proportion of Sarasota's pop is at higher risk of death, so it is not an
to comparison4/
In epidemiological terms, we would say that age is confounding our comparison of mortality rates across counties.
What is one way to make it a more fair comparison (at least on age)?
Something called "direct adjustment".
5/
What is one way to make it a more fair comparison (at least on age)?
Something called "direct adjustment".
5/
What we do is to pretend every county has the same age distribution.
Which age distribution do I use for each county? That is, what is the standard pop?
Well, no hard and fast rule here, but the overall statewide age distribution makes sense in this case.
6/
Which age distribution do I use for each county? That is, what is the standard pop?
Well, no hard and fast rule here, but the overall statewide age distribution makes sense in this case.
6/
I'll spare you the details, but what we do is to ask how many deaths a county WOULD HAVE HAD if it had the same population as the entire state. We call this the "expected" deaths.
From this, we calculate an "age-adjusted mortality rate".
7/
From this, we calculate an "age-adjusted mortality rate".
7/
These are not "real" rates, but do allow for a more fair comparison of mortality rates across counties with different age distributions.
Does it matter?
You betcha.
8/
Does it matter?
You betcha.
8/
Remember Sarasota (older) and Orange (younger) counties?
Of the 24 largest counties in FL:
- Sarasota has the 6th highest mortality rate (153.1 per 100k), which is 2.1 times higher than Orange (72.9 per 100k), which has the 2nd lowest rate
But when we adjust for age...
9/
Of the 24 largest counties in FL:
- Sarasota has the 6th highest mortality rate (153.1 per 100k), which is 2.1 times higher than Orange (72.9 per 100k), which has the 2nd lowest rate
But when we adjust for age...
9/
Then Sarasota County's age-adjusted mortality rate is 31% LOWER than Orange County's.
Orange is now in the middle of the pack (11th of 24), and Sarasota has the 3rd lowest.
10/
Orange is now in the middle of the pack (11th of 24), and Sarasota has the 3rd lowest.
10/
Leon County ranks 19/24 for the crude (unadjusted) mortality rate, but is in the top 5 highest when we adjust for age.
Why did Leon look so good?
Leon County is "young", with the 3rd lowest % of its population that is 65+ (among these 24 counties).
11/
Why did Leon look so good?
Leon County is "young", with the 3rd lowest % of its population that is 65+ (among these 24 counties).
11/
Sometimes it makes no difference. Top 2 highest rates are the same before and after age adjustment:
#1 Escambia
#2 Miami-Dade
But
rankings 3, 4, 5, and 6 before adjustment become
rankings 7, 9, 12, and 22 after adjustment.
12/
#1 Escambia
#2 Miami-Dade
But
rankings 3, 4, 5, and 6 before adjustment become
rankings 7, 9, 12, and 22 after adjustment.
12/
Clearly there are a myriad of other factors that muddy (i.e., confound) the comparisons made across counties.
Age is just the most obvious to illustrate.
Just be careful when you see county comparisons of metrics like mortality rates, or state comparisons for that matter.
13/
Age is just the most obvious to illustrate.
Just be careful when you see county comparisons of metrics like mortality rates, or state comparisons for that matter.
13/
Below are the final rankings for all 24 counties, before & after "age-adjustment". Remember, adjusted rates are for comparison only - since they estimate the rate IF THE COUNTY LOOKED LIKE THE STATE - they should not be used as "real" rates.
sorted by unadjusted rate.
14/end
sorted by unadjusted rate.14/end
