All the antibody studies in the US so far are showing that antibodies are present in at least an order of magnitude greater number of people than the official number of people with Covid-19.

This has huge implications for our response. Given this new data, LA is suggesting [1] that ~5% of the population of LA has already developed antibodies, which would mean the fatality rate may be as low as 0.10-0.20%. (For comparison, the seasonal flu is 0.05%-0.10%.)

Knowing if this is really the case for the entire nation would have huge implications for our response.

1. https://www.facebook.com/countyofla/videos/537241533852930/

- Identifying people who are carriers so they can be isolated

- Contact tracing (pointless at the moment, there are hundreds of thousands of people in the US who are infected)

- Ruling out other diseases or conditions that are treatable

Testing basically doesn't help anyone who has covid, it helps the people they would have spread it to if you can prevent that spread. There is no treatment for covid whatsoever, knowing you have it doesn't help you live longer or become healthy sooner. Knowing you don't have it, if you are already sick, might help if that lets them invest in further testing or treatments for other, non-covid, conditions.

Testing is easy to point at, but at the moment more testing would have limited impact. Even if we were to test everyone in the US, it would have to be done in a relatively short window of time to allow for a significant impact, and it's unlikely that there will be enough testing capacity to use this strategy before we reach saturation levels of infection. If they test me today and I'm negative, then they test you in 6 weeks, what does that get you and me? I was very likely infected in the meantime. This could be used socially like an negative AIDS test, except instead of a few months or years the validity of a test as a social currency is about a week, making it more or less useless as a way to avoid being infected.

* I would lead with the % of the population tested (not the total) and make it in the last 7 days, since I really care how likely they are to detect a current outbreak.

* I think to be really useful, it needs to be localized by county, since what I really care about is how likely I am to get infected, and that means I are about testing in my area.

* I would go with the positive rate in the last 7 days. It's important to explain that a high positive rate indicates they are probably only testing people with symptoms, and so they are probably missing lots of infectious asymptomatic cases.

* Then, I think I would drop the recent growth rate in testing, because I suspect that the growth rate is not going to grow consistently, and also that testing growth will be unevenly distributed. I am skeptical that it is meaningful to try to predict when testing will achieve specific milestones.

For epidemiological purposes, you only need a statistically valid sample, which is much smaller than "everyone".

For medical purposes, you only need to test people who are sick, which is hopefully also much smaller than "everyone".

In the us.

Nice.