I say that from the media-frustrated view of an American within America, who regularly reads German and Scandinavian news online (and UK too) just for external perspective.
I reflexively think two two things: (1) title sensationalism possible, and (2) or it could be far lower than reported, but that's not sensational, and only time will tell.
Systemic undercounts can happen easily in a situation like this; for example if you are only reporting as Covid-19 deaths people who a) died in a hospital and b) had a positive test. This will undercount if you are not testing broadly enough (or the test isn't accurate) or people are dying outside the hospital in unusually high rates due to the virus. NYC is a good example of where it's pretty clear this was happening (and they have adjusted their counts upward).
Fatalities is one area with overcounts are quite unlikely, but could be happening in some places I suppose, and will have to be accounted for if it can be demonstrated.
Somebody who tests positive, and then dies in a car accident meets your above criteria. For something less contrived, so does somebody who tests positive, and then subsequently dies of an unrelated heart disease.
If you wanted to fix attribution, you’d need a coroner’s investigation into each death, which certainly isn’t happening. Even testing every death that you want to count isn’t feasible, because of test shortages and the potential for false negatives. A real world example of that is in New York, where a revision in their attribution methodology had their numbers jump up massively in one day [0]. What changed was they decided to add all deaths where Covid was a suspected cause to the count.
Whether these shortcomings in methodology would lead to an overall overcount is probably unlikely. But methodology differences from one place to the next does make it difficult to compare statistics. With some factors applying downward pressure to the stats, and other factors applying upward pressure.
One thing I’m going to be interested in looking up when this is all over is 2020s influenza deaths. I have a strong suspicion that 2020 is going to have record low flu deaths, with many of them being incorrectly attributed to Covid-19.
[0]: https://www.theguardian.com/us-news/2020/apr/15/new-york-cit...
[0]: https://www.nytimes.com/interactive/2020/04/21/world/coronav...
I agree that over time it gets muddier , but right now in places like NYC it’s pretty clear .
New York pretty clearly did the right thing here. There isn't going to be a coroners investigation, but with somewhere that large you have decent statistics on expected rates and attributing the up tick they saw was methodologically better than excluding those deaths. After all if I understand it correctly "suspected deaths" here is counting the otherwise unaccounted for increase over baseline statistics, not the baseline cases. That's the correct way to do this, and afiak it is how NY did it in reporting but I haven't double checked. It's not perfect, but that's life.
Agree that this years flu analysis will be interesting, but on the overall statistics nobody competent is going to conflate them with covid-19, there may just be larger error bars.
> with somewhere that large you have decent statistics on expected rates
I wouldn’t make that assumption. I can’t think of any place in the world that has the ability to collect this data with any sort of scientific rigour. It’s simply not possible at the moment, and without that you have no scientific basis to know what to expect.
> Agree that this years flu analysis will be interesting, but on the overall statistics nobody competent is going to conflate them with covid-19
I wouldn’t make this assumption either. If you attribution criteria includes dying in hospital of Covid-like symptoms, then it also includes dying in hospital of flu-like symptoms, because a severe case of flu and a severe case of Covid present almost identical symptoms. The US has had ~55k Covid attributed deaths so far, in last years flu season, it had ~35k flu attributed deaths. The potential for misattribution on that basis alone is significant.
The general point I was making was that not all of the confounding factors apply downward pressure to the stats, some of the provide upward pressure, and depending on methodology, you’re going to get a different set of confounding factors from place to place.
I’m disagreeing that at this moment - and with statistics like excess mortality in NYC, there is very good reason to believe that upward and downward pressure are not equally likely.
As to the first assumption I disagree . We have pretty good ideas of all causes rates in somewhere like NYC, we have pretty good numbers on things like accidents etc. And then sort of factors you point out like the effect of a recession aren’t on effect yet for the most part. So you have a bunch of excess morbidity, and you have pretty good local statistics in the gaps, at least locally. You are right that this doesn’t go apples to apples in other places but we aren’t doing that so it does matter.
There is one kind of interesting wrinkle, you may have stress related impacts that show up even on a short timeline. So there is a question if you can/should attribute those and you can pick reasonable ways, either choice .
Flu isn’t likely going to be confounded too badly because people are going to be focussing in the counfounding impacts (e.g. of isolation). It just seems strange to me to assume this will be a problem , some people involved are good at this sort of analysis . Much more likely that a decent job will be done and the estimation errors will have wider bounds.
I guess I just don’t see any data to support your pessimism , on that , but I could be wrong of course. I mean this sort of modelling isn’t easy but it is doable.
I would kinda agree, but for different reasons. I’ve seen the inner workings over several government agencies during this crisis, including a few police departments. What has been obvious to me is that there is a very significant difference from place to place in how death and infection statistics are recorded. My observation has been that people who have an obvious political incentive to record lower numbers usually find a way to do it, and vice versa for people who have an obvious political incentive to record higher numbers. I would agree because across the world, there seems to be greater political incentive to have lower numbers.
Everything you’re speculating about is plausible, and may turn out to be mostly correct. But there is no reasonable basis for the certainty you claim to have. Especially in regards to claims such as there being no potential for factors other than Covid deaths to have a statistically significant influence on overall mortality rates. That’s just your opinion, and it’s a reasonable on to have, but it’s definitely not “almost certainly” true.
>The death toll from coronavirus may be almost 60 per cent higher than reported in official counts, according to an FT analysis of overall fatalities during the pandemic in 14 countries.
The article is about reported deaths, not predicted. I doubt many countries are reporting artificially high counts.
To calculate excess deaths, the FT has compared deaths from _all causes_ in the weeks of a location’s outbreak in March and April 2020 to the average for the same period between 2015 and 2019.
so they are not counting only coronavirus related deaths but all
So naturally people get cheesed off. One side has been hit hard, another hasn't and both get grouped together to produce some arbit number.
Rather than counting country wise what would make more sense is global hotspot comparisons(town/city level).
The hotspots will see a lot of Excess Deaths over historic mortality rates I have no doubt.