[sleavey]

It's worth noting that the risk of death stated for AstraZenica in this article, one in two million, surely relies on proper and rigorous reporting of the causes of deaths. As a thought experiment, let's say someone aged 65 happens to die one month after having received the first dose of the vaccine. Are we sure that the local coroner will check whether the vaccine had anything to do with their death if there is some more "obvious" cause of death like an underlying health condition? Are we sure to the level of one in two million? It would only take 18 misattributed deaths to make that risk 10x higher.

I'm not trying to start an argument over vaccines but it sincerely seems to me like there might be large error bars on the assertion that the risk of death from the vaccine is one in two million. And I expect aspirin is still far riskier than the vaccine so the central headline doesn't change. I'm probably missing something here so would welcome input from others. Hopefully Hacker News is a mature enough place to discuss this with level heads.

EDIT: 2 minutes after posting, first downvote. Maybe my previous sentence was too hopeful?

[CorrectHorseBat]

And what about the other way around? What is the chance that a 65 year old randomly dies of something that could be misattributed to the covid vaccine. It would only take 2 misattributions for the risk to be zero.

[sleavey]

Indeed, and that doesn't change my point (though I'm not accusing you of saying that it should). Do we believe accuracy in such accounting - at least in Australia - is error-proof to at least one in a million? Could it be 10x better or worse? 100x? Where's the line? I actually have no idea where you would even begin to work out the accuracy of such a number. For aspirin too, for that matter.

[th0ma5]

Yes it is very true that not very many people understand risk enough to effectively compare it amongst other risks.

[splistud]

And it is also true that no one can effectively compare risks when the assumptions can not yet be relied upon