This is strongly reminiscent of https://www.buzzfeednews.com/article/nidhiprakash/coronaviru... that anecdotally reported that COVID deaths are being severely underreported in the USA because doctors were told to only test if it would change the care given because tests were in short supply. But that means that if the person goes on to die, they are not actually counted as a case.
The article doesn't explicitly make the point, I think a lot of the relevance of this potential undercounting is how it relates to assessing preventative measures. Many people have argued (correctly) that "confirmed cases" should be ignored, as the numbers produced are so dependent on non-random sampling. But they often followup by arguing that we should focus on COVID deaths, because those are (usually) too hard to hide. Potential undercounts of this sort call that conclusion into question.
Perhaps paradoxically, going up another level and looking at "all deaths regardless of cause" doesn't necessarily improve the quality of the decision making. If anti-COVID measures also serve to decrease flu deaths and traffic accidents, it's possible that the total numbers might show a drop even if those same measures are inadequate for reducing COVID deaths. It's hard to get anything right!
But they often followup by arguing that we should focus on COVID deaths, because those are (usually) too hard to hide. Potential undercounts of this sort call that conclusion into question.
Which was the point of both the OP and the buzzfeed article that I pointed.
We only count someone as a case when they get tested and are positive. However we don't test someone simply because they are really sick and it looks like COVID-19. Which means that nobody has to hide the body - it simply isn't counted in the COVID-19 numbers.
This systemic source of error seems to be common across many regions and countries. And it throws all quoted statistics into question. And not by a handful, but by a significant factor.
Of course at the moment preventing bodies is more important than counting them. But knowing that we can't trust the counts either, matters.
Perhaps paradoxically, going up another level and looking at "all deaths regardless of cause" doesn't necessarily improve the quality of the decision making. If anti-COVID measures also serve to decrease flu deaths and traffic accidents, it's possible that the total numbers might show a drop even if those same measures are inadequate for reducing COVID deaths. It's hard to get anything right!
True. But if the systemic undercounting of COVID-19 deaths is a reasonably large factor and the death rate is a significant fraction of the total death rate, it can become a more accurate estimate than the official numbers.
And yes. Accurate decision making with unreliable numbers is impossible. The best we can do is heuristics.
> Which was the point of both the OP and the buzzfeed article that I pointed.
Sorry if my comment was clumsy. I realize that you are aware of everything I said. I didn't intend it to be novel. Trying again:
On a superficial level, there's been a lot of media coverage of the number of confirmed infections (the denominator). On a slightly more advanced level, there have been calls to pay more attention to the actual number of deaths (the numerator). These discussions properly point out that deaths (which can theoretically be counted) are likely to be more accurate than infection numbers (which can only be estimated).
This article (and the one you linked) are pointing out that that the number of deaths due to COVID-19 also has significant uncertainty. My comment was trying to make clear why this matters: not for you, but for others who have been paying attention mostly to the media coverage. Perhaps this 538 article makes the point better:
https://fivethirtyeight.com/features/why-its-so-freaking-har... (https://news.ycombinator.com/item?id=22746780).
reply