Hacker News Reader

> Giesecke has claimed less than two weeks ago that at least 600k people in Stockholm (pop 950k) have had it

Stockholm has 2.34 million inhabitants. (Stockholm in this context is always stockholm area, not the city). That makes a 600k estimate much less outlandish although still optimistic (25%). Self-selecting hospital workers showed 20% in yesterdays publication, which seems like it should be an upper bound for the general population.

It's also important to remember that when people make the claim that "X had it" they also simultaneously estimate that perhaps X/2 would show up as positive in serological tests (because the rate of infection and delay of antibodies would create a lag) so a person claiming 20% having been infected isn't contradicted by a serological result of 10% positive.

> Stockholm has 2.34 million inhabitants. (Stockholm in this context is always stockholm area, not the city).

I agree, counting only Stockholm city isn't reasonable. But I've found the 'Metropolitan Stockholm' number is seldom used as well, but looking at the definitions it should be more common.

What is more commonly used is the number for the Stockholm urban area (I'll link the Swedish wiki article since it contains a lot more data, https://sv.wikipedia.org/wiki/Stockholm_(tätort) ).

It's interesting to compare the metropolitan areas. In Stockholm is stretches so far that it includes areas that feed to (and somewhat off) some of the other large cities (mostly Uppsala I guess), but in Skåne the metropolitan area only covers a few municipalities although commuting by car is more commonplace and there seems to be a lot more commuting by train from Helsingborg and Kristianstad to Malmö (it's also extremely handy by train) than from Uppsala to Stockolm (no hard numbers, but I commuted Uppsala -> Stockholm for five years and Helsingborg -> Malmö for a while).

So it does make more sense to use the 'Greater' areas for all the major cities.

> But I've found the 'Metropolitan Stockholm' number is seldom used as well, but looking at the definitions it should be more common.

The county (region, or formerly "län") is the area responsible for healthcare so it's natural that this area is used for all things healthcare related.

https://en.wikipedia.org/wiki/Regions_of_Sweden

https://en.wikipedia.org/wiki/Stockholm_County

Yes. For Stockholm "Stockholmare" is used for multiplex definitions when for Malmö you have "Malmöiter" and "Skåningar" where the first is specifically the urban area and the latter for the region (healthcare) but an area greater than the metropolitan area.

Where does Giesecke propose a strategy of heard immunity? Does every set of restrictions that leave out total lockdown count as a "herd immunity" strategy?

I only see one of these three options:

1. He's aiming for herd immunity, using the people who are the less likely to have severe cases to protect the most vulnerable

2. He believes the death toll will really not be that high to justify the actions that we are taking to prevent the virus from spreading

3. He does not care about the deaths, and the faster they die, the faster we can go back to normal

Of course I don't think he believes in number 3, which is quite horrible in my opinion.

Regarding 2, the numbers emerging of the population that has actually contracted the virus seem to be much smaller than what Giesecke was assuming. So it would seem correct to assume that the death toll is going to be much higher than he was anticipating based on a wrong hypothesis.

To me it only leaves number 1 as a potentially valid approach that does not rely on quarantine.

Thanks for clarifying that your attribution of this strategy to Giesecke is entirely your own invention.

It wouldn't make much sense to use population who have antibodies as the main goal, and view the peak as desired but reached without further restrictions because of increased resistance in thebpopulation as an argument otherwise. He isn't arguing for lowering infection rates as long as healthcare holds, but for increased immunity. That's not something to be mad about.

Folkhälsomyndigheten has now spent over a month outright denying that the strategy is aiming for herd immunity: the strategy is purely dedicated to keeping the pressure on the health services manageable. This isn't really an issue about where you personally think it's the correct strategy, rather it's a about whether you believe what the authorities tell you.

The most credible connection to FHM this strategy has is that a former employee wrote a mail explaining the reasoning of not closing the schools to a friend who is a politician working on local school issues, the friend then (with permission) shared the text further on FB.

> Den tidigare statsepidemiologen Annika Linde hade inte tänkt ge sig in i diskussionen om Folkhälsomyndighetens strategi för att bekämpa det nya coronaviruset. > > – Jag halkade in på ett bananskal, säger hon. > > Bananskalet var en fråga den 12 mars från en lokal skolpolitiker som ville ha hjälp att förklara varför det är rätt av Sverige att inte stänga skolorna. > > Varför säger inte Folkhälsomyndigheten bara att flockimmunitet är planen, undrar folk i sociala medier upprört efter Lindes inlägg. Andra tycker att planen låter rimlig. > > – Sedan förnekade Folkhälsomyndigheten att det var tänkt så, säger Annika Linde. > [...] > > Den tidigare statsepidemiologen Annika Linde tänker inte så mycket på att hennes budskap om den svenska strategin tog sig hela vägen till Vita huset, säger hon. Däremot tänker hon fortfarande att flockimmunitet är målet för Sverige, i väntan på ett vaccin. Men hon försöker låta bli att kommentera sina efterträdares arbete. > > – Det finns något som kallas ”lösa kanoner på däck”. De skjuter lite hit och dit och kan träffa skeppet och sänka det helt och hållet och sedan inte ha något ansvarstagande. Så jag nöjer mig numera med att svara på enstaka frågor, som nu från dig om att jag inte höll tyst där i början. https://www.dn.se/nyheter/sverigebilden-under-coronakrisen/

> A large anti body study was supposed to be released yesterday, and I'm waiting to see what it says.

Also be aware that there are false positives, especially significant with these antibody tests, so with the low numbers of positives it's extremely important to evaluate if the claimed values are more than noise artifacts of the tests themselves. It is also important to be aware of the scenarios for which the use of the apparent test results is not reasonable.

Moreover, here's what happened in the UK a few weeks ago:

https://www.theguardian.com/world/2020/apr/09/uk-government-...

"None of 3.5m home tests ordered have so far been accurate enough to detect coronavirus immunity"

UK got 3.5 million(!) unusable antibody tests.

https://www.bmj.com/content/369/bmj.m1449

"John Newton, Public Health England’s director of health improvement, said:"

"A number of companies were offering us these quick antibody tests, and we were hoping that they’d be fit for purpose, but when they got to test, they all worked but were just not good enough to rely on.

“The judgment was made [that] it’s worth taking the time to develop a better antibody test before rolling it out, and that is what the current plan is.”"

"Newton told the committee that the tests trialled so far had lacked sufficient sensitivity to identify people who had been infected. “We set a clear target for tests to achieve, and none of them frankly were close.”"

I also share your concern about claims made by Gieseke.

The article quotes Gieseke as saying "The real death toll, he suggested, will be in the region of a severe influenza season — maybe double that at most".

Sweden's official death toll was 2270 yesterday (2020-04-28). The three worst influenza seasons in Sweden since 1969 killed 807, 674 and 652 people[1]. So we've already passed his "at most" claim. The only way I can see that working out is if he feels that none of the influenza seasons in the past 50 years count as "severe".

[1]: numbers taken from the Swedish wikipedia entry on influenza. I didn't bother to check wikipedia's source. An average Swedish influenza season kills about 200.

> he only way I can see that working out is if he feels that none of the influenza seasons in the past 50 years count as "severe".

Or that he doesn't think deaths in past influenza seasons were all attributed to influenza but rather just "normal" deaths.

The excess all-cause deaths will be the figure to look at, but it will take quite a while before those numbers are reliable.

But the excess death data will also be tainted because people, especially older people, avoid going to hospitals out of fear of catching covid19. Cancer screenings are way down, and fewer people go to the ER with heart attacks and the like. This will inevitably lead to excess mortality down the road --even among otherwise healthy people--, but it will be very hard to determine the magnitude of this second-order effect.

Of course it's also very hard to distinguish between people dying with covid19 and dying because of it. And we've never even made a serious effort to track the cause of death of the elderly.

Meaning we don't know how many people die each year from the flu, and we don't know how many people are dying of covid19. When you add it all up, it will be very difficult to learn the right lessons from this pandemic.

That's not what excess mortality means. They don't say "this many extra people died, I guess it must be covid" -- they're not fucking idiots.

When cancer patients die because their cancer treatment was cancelled they die of a cancer related cause, and that's how their death will be recorded, and that's how their death will be reported.

> And we've never even made a serious effort to track the cause of death of the elderly.

It's hard to understand your "we" here. Which country doesn't try to track cause of death for elderly people?

> Meaning we don't know how many people die each year from the flu,

But we can count the deaths the same way. We can look at deaths of people confirmed to have the disease, we can look at death certificates, we can look at excess mortality combined with community surveillance. The errors for all three are going to be similar for flu and covid-19.

> But we can count the deaths the same way.

When person in an elderly care facility dies we don't do a forensic investigation. We just shrug and say "I guess it was their time". So we don't have an accurate mortality baseline to do any comparison against. And that's assuming we are accurately distinguishing between those who died of covid19 and those who died with covid19, which we don't.

I'm not suggesting that people who draw conclusions from incomplete data are idiots, I'm pointing out that the data we have is completely insufficient to make an accurate assessment of covid19 mortality.

I think this is why excess all cause deaths will be the best evaluation. Lots of people who die from Covid in care homes are people who would have died from this years flu anyway. They aren't excess deaths, they are the baseline deaths.

I agree that 'excess all cause deaths' is an important metric, but we still have to correct for clear biases in both directions. E.g. fewer traffic deaths because of a lockdown.

> When person in an elderly care facility dies we don't do a forensic investigation. We just shrug and say "I guess it was their time".

This ("we just shrug and say "I guess it was their time"") is untrue. It's okay that you don't know, but you should stop spreading this misinformation.

The first link I get when I google "accuracy of death certificates" is this one:

The accuracy of death certificates. Implications for health statistics. https://www.ncbi.nlm.nih.gov/pubmed/1871957

> Significant discrepancies between the two documents were observed in 50% of patients. In 25%, the immediate cause of death was incorrectly stated on the certificate, having been assigned to a different organ system in the majority of those cases. In 33%, there was disagreement on major disease other than the immediate cause of death.

Anecdotally, elderly people I've lost had completely wrong death certificates. When common sense, anecdotal evidence, and a cursory review of the scientific literature point in the same direction I'm going to assume that's the way it is.

Also, try to be kinder in the way you communicate.

This is what's concerning all over the world.

We know that death rates with the disease vary a lot due to attribution.

We really can't get a good comparison as influenza isn't usually attributed as the main cause of death but coronavirus is.

We can only look at all cause mortality rates which is quite the lagging indicator.

https://www.euromomo.eu/graphs-and-maps/

The best we have.

> as influenza isn't usually attributed as the main cause of death but coronavirus is.

What makes you say this?

The UK made SARS-CoV-2 a "notifiable disease" which means by law any death from it must be reported to central government and it must be allocated to that death when present.

https://www.spectator.co.uk/article/The-evidence-on-Covid-19...

But there’s another, potentially even more serious problem: the way that deaths are recorded. If someone dies of a respiratory infection in the UK, the specific cause of the infection is not usually recorded, unless the illness is a rare ‘notifiable disease’. So the vast majority of respiratory deaths in the UK are recorded as bronchopneumonia, pneumonia, old age or a similar designation. We don’t really test for flu, or other seasonal infections. If the patient has, say, cancer, motor neurone disease or another serious disease, this will be recorded as the cause of death, even if the final illness was a respiratory infection. This means UK certifications normally under-record deaths due to respiratory infections.

Now look at what has happened since the emergence of Covid-19. The list of notifiable diseases has been updated. This list — as well as containing smallpox (which has been extinct for many years) and conditions such as anthrax, brucellosis, plague and rabies (which most UK doctors will never see in their entire careers) — has now been amended to include Covid-19. But not flu. That means every positive test for Covid-19 must be notified, in a way that it just would not be for flu or most other infections.

The Spectator gets almost all of that wrong.

1) Death certificates rely on doctors using their best knowledge and experience to say what the patient died of, and what the patient died with. Flu and other respiratory illness is mentioned on many death certificates. See 5.4 here: https://assets.publishing.service.gov.uk/government/uploads/...

2) A notifiable illness has no meaning for death certificates.

3) We don't routinely test for flu. That's why all cause mortality is the preferred statistic for flu deaths, and also for covid-19 deaths.

Huh. That's not the response I expected.

You're not arguing with the Spectator, they're just acting as a publisher. You're arguing with in his words, "a recently-retired Professor of Pathology and NHS consultant pathologist". So you're claiming a professional British pathologist doesn't understand how British death certificates work, and you know better. Bold move.

Especially so because you seem to actually be agreeing with what he wrote, which is weird. For instance your claim (3) exactly matches his claim that:

"the vast majority of respiratory deaths in the UK are recorded as bronchopneumonia, pneumonia, old age or a similar designation. We don’t really test for flu, or other seasonal infections."

You say flu appears on many death certificates. Yes and he never argued otherwise. He said despite that it's sometimes mentioned it's actually under-reported because testing isn't really done much for it - as you agreed with!

That leaves the question of notifiability. The rules say that COVID-19 must be mentioned on a death certificate if testing was done at all, even if negative (which is new to me, I wonder what that does to the widely cited stat of "number of certificates that mention COVID"). But the point is that relative to flu, testing deaths for COVID is enormous, practically blanket at this point. If you test a lot and you insist that every case is reported to central government it will cause a flood of reports to arrive on the desks of decision makers, who will then feel it's much worse than flu. But it's not, it's just a reporting artifact.

During the last four influenza seasons the average deaths was 685 and the worst season was 2017−2018 when a bit more than 1100 people died.

Source: https://www.socialstyrelsen.se/globalassets/sharepoint-dokum...

That difference seems to arise because the report you cite uses a broad definition of influenza deaths. Page 2 says they're counting patients who "have influenza as a contributing or underlying cause of death, plus patients who received specialist care with an influenza diagnosis who subsequently died within 30 days".

In any case, boosting the number to 685 still isn't enough so that covid is "double at most", and boosting it to 1100 isn't enough either.

As far as I know they have an even broader definition of corona deaths in Sweden.

My point is more that statistics is hard. You tried to debunk someones "dodgy" statistics with your own numbers that can be twisted to be even more off than the person you tried dismiss. I'm afraid nobody and everybody will be right after this is over. People will always find statistics to prove their points and that they where right and the other people where wrong.

> Sweden's official death toll was 2270 yesterday (2020-04-28). The three worst influenza seasons in Sweden since 1969 killed 807, 674 and 652 people[1]. So we've already passed his "at most" claim. The only way I can see that working out is if he feels that none of the influenza seasons in the past 50 years count as "severe".

> [1]: numbers taken from the Swedish wikipedia entry on influenza. I didn't bother to check wikipedia's source. An average Swedish influenza season kills about 200.

Sorry, but you can't just throw around numbers without explaining exactly what they mean or where they come from. The numbers you're quoting is most likely deaths that has been diagnozed as influenza. To get the full picture you need to look at excess mortality (which is reported by EuroMOMO[1]) and possibly adjust the numbers to pick out the influenza-related excess (FluMOMO[2] is the model most countries use).

If you look at the 2016/2017 season in Sweden [3, figure 17, page 46] you will see that the excess mortality as reported by FluMOMO goes way beyond ~600 for a season. In the peak season we see that it was ~300 per week. There are of course uncertainties in these numbers (which is why you won't see any official "x number of people died of influenza" figures), but it was probably closer to thousands than hundreds in 2016/2017.

[1]: https://www.euromomo.eu/ [2]: https://www.euromomo.eu/how-it-works/flumomo [3]: https://www.folkhalsomyndigheten.se/publicerat-material/publ...

> you can't just throw around numbers without explaining exactly what they mean or where they come from.

In the interview, Gieseke says influenza kills 1000 to 2000 people per year in Sweden, it's part of the exchange starting at 24:16. He doesn't explain exactly what those numbers mean or where they come from.

A minute or two later, he guards his 'double' comment by saying it's not going to be 10x.

One interpretation of the exchange is that he's predicting 2000-4000 deaths in Sweden, and definitely not 20000.