Lies, Damned Lies, and Statistics

There are three degrees of untruth, – a fib, a lie, and statistics. – Sir Charles Dilke (1891)

A recent NYT article titled "In Florida, the pandemic is worse now than it has ever been before" opens with a bang:

More people in Florida are catching the coronavirus, being hospitalized and dying of Covid-19 now than at any previous point in the pandemic, underscoring the perils of limiting public health measures as the Delta variant rips through the state.

The article continues, highlighting that raw death counts and hospitalizations have indeed increased across Florida.

Not awesome.

And it's not a data integrity issue since the data triangulate with what we're seeing when we look at the excellent COVID tracker courtesy of the Financial Times:

The key thing to notice about this graph is again, we are looking at new confirmed cases which are not indexed to the population size (in case you were wondering, I've listed the top states by population estimate).

1. California (39,512,223)
2. Texas (28,995,881)
3. Florida (21,477,737)
4. New York (19,453,561)
5. Pennsylvania (12,801,989)
6. Illinois (12,671,821)

It's notable that both TX and Florida are smaller states and have a higher 7-day rolling average of new confirmed COVID cases, so we probably want to dig a bit deeper to better understand the dynamics at play.

First, let's look at new confirmed cases per 100k which provides a much better way to compare across states which, as mentioned above, have differing populations. When we apply this filter, things still don't look great for Florida and Texas.

But should we be looking at new confirmed cases of SARS-CoV-2? COVID cases do matter and all else equal, I'd rather not have COVID than have COVID and COVID infections lead hospitalization and death. That said, I'd submit that deaths attributed to COVID per 100k over the totality of the pandemic is the most straightforward way to measure pandemic responses across various jurisdictions. More importantly, it's the metric we should care the most about since death is the worst outcome one could experience.

Looking at new deaths attributable to COVID-19 in the 6 largest states, we see that TX and FL are right at the top (though, for some reason CA stopped providing data on August 8th).  

Here's where things get interesting. If we look at cumulative deaths per 100k over the course of the entire pandemic, we can start to see how different public health regimes have varied across states. New York, for example, started fast out of the gate with the state mandate that nursing homes accept COVID-19 positive patients and is a large reason for why they reached 160 deaths per 100k almost 8 months before Pennsylvania and Illinois hit that mark.

What we see is that despite Florida's uptick in both COVID cases as well as COVID deaths, their overall trajectory in terms of cumulative COVID-related deaths is still better than states like New York, Pennsylvania, and Illinois.

Here I've unhid some of the unselected states so we can get a feel for where Florida falls relative to the rest of the country.

Adding in a slightly different mix of states, we can start to understand why the NYT article cited at the beginning is so deceiving.

Yes, it's true that Florida has spiking rates of COVID infection, hospitalization, and death. No, that is not good. That said, relative to Florida the rate of cumulative death per 100k is 34.2% higher in Massachusetts, 41.48% higher in New York, and 53.38% higher in New Jersey (as of August 25th, 2021).

Circling back to the NYT and the article that prompted this post, I'll just leave you with is this.