undefined | Better HN

0 pointstheptip5y ago0 comments

Got you, thanks. (Side-note, I wish it were easier to transmit simple models like this as part of discourse).

The 0.0015 risk of death seems high. Where are you getting that from?

https://www.reuters.com/article/us-health-coronavirus-german... gives "The EMA has said that as of March 10, a total of 30 cases of blood clotting had been reported among close to 5 million people vaccinated with the AstraZeneca shot in the European Economic Area, which links 30 European countries."

But if you want to take the German numbers, 3 deaths out of 1.6m vaccines:

3/1,600,000 = 0.000001875 (In percent, 0.0001875%) which is 0.125 of your figure.

And the OP gives a base rate of "two to five cases per 1 million individuals per year" for context; we need to subtract out the base rate from the observed deaths too (because obviously the base rate of this condition occurs regardless of whether you take the vaccine).

[edit: percent conversion typo]

0 comments

petre5y ago

There's a mistake on my part, the risk of getting blood clots, not death, is actually even lower, 30/5/1e-6 = 6·1e-6, not 0.0015.

The base rate is per year, the AZ vaccine was used for about 3 mo in the EU. That leaves us with a base rate of 0.5 to 1.25 cases per million per 3 mo, so using your numbers:

1.875 - (0.5…1.25) = 1±0.375

deaths per million if we were to keep the inputs unchanged?

theptipOP5y ago

I knocked together a quick Collab for this; feel free to clone and work with it.

https://colab.research.google.com/drive/1C7K3u3vKrIMt-sa4_t3...

The first cell is the model I was building, I finally got it into reasonable shape. It comes from the published national statistics for death rates. (Comments welcome, I just sketched this to help me think about the numbers here. No claim on this actually being right.)

The second cell is my attempt at rendering your initial calc, though I've not folded in the base rate changes. We can go into the weeds there but looking at the results I don't think we need to; my calcs naively attribute all the blood clot cases to the vaccine and still give COVID as being worse than the vaccinations by a factor of 400-800.

The way you're thinking about base rates sounds right to me though, and it does sound like the best guess is we're at something like 4x above base rate.

Given that I implemented my model with a completely different approach and we're both within a factor of 2 (if I've rendered yours correctly) I think we're in the right ballpark here, at least for "Fermi calculation" level of completeness.

Guardian article with more numbers: https://www.theguardian.com/world/2021/mar/16/benefits-of-as...

j / k navigate · click thread line to collapse