Thread by @youyanggu

Florida either has an incredibly low fatality rate (IFR), or may be underreporting deaths.

I computed the implied IFR for every state based on my analysis of reported cases, test positivity rate, and reported deaths.

According to my calculations, Florida's implied IFR is 0.15%.

Here are the implied IFR for all states with more than 1,000 cases/day this past week (chart above). The implied IFR for US as a whole is currently 0.27%.

0.15%: FL
0.16-0.20%: GA, NV, TN, LA, AZ
0.21-0.25%: NC, TX, SC, CA
0.26%-0.30%: AL, MS
0.31-0.45%: OH, IL

For states with lower than expected implied IFR, I can come up with a few hypotheses:

- Better treatment
- Lower median infection age
- Lower test positivity than reported
- Significant delay in death reporting
- Underreporting of deaths

Let's take a look at Florida.

Based on their 20% test positivity rate, I estimate that true infections is ~10x higher than reported cases (will explain more in future posts).

That corresponds to ~100k true infections/day in late June, or a 0.15% IFR after adjusting for lag.

We can plot the implied IFR over time, and you can see that the implied IFR for Florida went from 1% in May to 0.15% in late June. In the same time span, the US overall implied IFR went from 0.65% to 0.27%.

Interestingly, 0.65% is the US IFR value currently cited by the CDC.

It seems unlikely that Florida would have significantly better treatment or a lower median infection age than the rest of the US.

If we assume a 0.3% true IFR, that would correspond to ~300 deaths/day. So the fact that they are only reporting 100-150 deaths/day is unusual.

If Florida's reported deaths do not significantly increase, then it's possible that we'd see ~1,000 unexplained excess deaths per week in July.

Florida's the most extreme example, but other states like Georgia, Nevada, Arizona also have very low implied IFR.

This is just some preliminary results. I am currently working on better characterizing the relationship between reported cases, positivity rate, and IFR. Will share more soon.

This is separate from my modeling work, which only uses past deaths to predict future reported deaths.