The Good News (for NY) and the Bad News (for Utah)

The big news today is that the state of New York did a carefully designed, randomized, serological  study.

Serology measures the number of people who have been infected with something and have developed antibodies to it.

The data was startling: the point estimates are 14% of people in the state, and 21% within New York City itself have had COVID-19. I have yet to see confidence intervals for that, but with a sample size of 3,000, I'd expect them to be about ± 2%.

The good news, which didn't require much deep thinking, was immediately picked up on in the legacy media: if a lot more people have had the disease than we thought, then its fatality rate is a lot lower too.

The bad news for Utah (and places like it) comes in two parts: one that is bad, and one that's neutral but sounds bad.

The bad one is that if the fatality rate is lower than we thought, then we can reverse the logic and say that places that have not done the serology work probably have a lot more undiagnosed cases than we thought. So COVID-19 is probably more widespread in Utah than it seems.

The more neutral one begins with the idea that we all knew (even though it is hard to admit) that with no immunity, a large fraction of the population of the planet is eventually going to get COVID-19. We now know New York state and New York City are much further along that path than we thought they were. And it makes sense that Utah is too: that's the neutral part. But it doesn't change the fact that we may still have to experience what New York City has gone through: they didn't get to 21% of the population having the disease without mass trench burials.

The important upshot from these statistics is that they will probably be used by policymakers to argue that COVID-19 is not as dangerous as we thought (which is true), but it does not change the fact that any particular location may have to run the gauntlet of a full-blown outbreak.

Let's crunch some numbers.

• New York City currently has 1,727 positive cases per 100K people. 
• But the serology indicate 21,200 people have had COVID-19. 
• So testing has found less than 10% of them, or alternatively that infections are 12 times cases. 
• For Utah that means that while we have 3.5K positive test results, we probably have 42K people who have had the disease. (That's 500 in Iron County, and 2K in Washington County).
• And those 3.5K positives work out to about 11 per 10K. 
• Worst case, our outbreak needs to double 6-7 times to get to the scale of the one in New York City. We had to double 11-12 times already to get from that first initial case to the 3.5K we have now. 
• Our current fatality rate is about 0.1 per 10K people. That's great. But doubling 6 times gets that to 12, and 7 times gets it to 24. That's a much better rate than New York City's 129 per 10K (probably reflecting that we live healthier, and we're more spread out here). But it still gets us to 400-800 deaths in the state (and 5-10 in Iron County, and 20-40 in Washington County).

So here's some food for thought: if our outbreak doubled 11 times already, do we really think it's not going to double 17 times in total? I put it to you that it's magical thinking that it can't.

But it might not. Maybe, just maybe, we are social distancing enough to get R0 down near 1, and we're persistent enough to keep at it. Maybe. May is not going to be fun. Stay safe.


† I lived my first 25 years in New York state. A problem with analyzing county level data is how to deal with Long Island. People in New York know how to do these stats properly, but I'm never sure when I see something like this in national media. Anyway, Long Island has 4 counties on it (Kings, Queens, Nassau, and Suffolk). But Kings and Queens are also in New York City. So when someone refers to Long Island, you need to figure out if it's the whole island, or just the suburban parts. I couldn't tell from the sources I used, so I didn't go there with deeper calculations.