Hacker News Reader

I'd also argue that looking solely at excess deaths is a rather narrow view of how Sweden's strategy has panned out, given what we're learning of "Long Covid"'s[1] impacts on the human body.

It's also distorted by accessibility to quality healthcare etc. Covid caused ICU bed nights would be an interesting metric, but again, distorted by that accessibility issue.

Indeed, I'd suggest looking at excess deaths is perhaps misleading for other countries like the US due to the access to care factor. Excess deaths ends up acting both as a measure of infection, and a critique of healthcare policies.

Something like deaths / cases would be interesting, but then you're introducing the variable of "how many cases were actually discovered via testing". I don't think I need to link to the head of the US government's claims that increased testing only causes rates to go up...

[1]: https://www.theguardian.com/world/2020/nov/15/damage-to-mult...

The nice thing about looking at excess deaths is the specificity and finality of the metric.

“Case” is problematic due to testing errors and sampling bias (and the fact that “case” in the medical sense generally means you are under a doctor’s care).

“Hospitalization” may mean “with COVID”, to manage hospital resources for infection control, or “from COVID”, the popular interpretation to assess how serious the illness is and how quickly the serious illness is spreading.

We can all agree on what death means. There is still the forensic issue of untangling how much COVID contributed, and that has been a focus of CDC work. But it’s a cleaner metric than most. Another post here gives the CDC site link, but here it is too (gives State-by-State info as well)

https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm

Might be nice to have a QALY-based assessment at autopsy...how many more good years would this person have had without the COVID-associated event? Admittedly a guess, but insightful nonetheless.

https://en.wikipedia.org/wiki/Quality-adjusted_life_year

One of the major impacts of CVOID is likely going to be on men's fertility.

> And further: looking at US figures for 2017, November tends to have about an average number of daily deaths, but December is one of the deadliest months, with about 15% more deaths-per-day than is typical in the lowest-deaths Summer months of June-July-August.

> So even my extrapolation to 90,721 is likely to be a floor on the true end-of year number, if lagged reporting or a typical December death acceleration means these November 6 numbers aren't an accurate sample for the full year.

I don't think you can assume that. While November to February are usually the months with the most mortalities, it is not certain that pattern will hold this year. The winter months kills a lot of vulnerable people because it is cold outside so people remain indoors close to each other causing infectuous diseases to spread. But this year thousands of vulnerable people died in April and May and there may not be that many left for viruses to kill, so to speak.

Here is some interesting statistics showing that September 2020 was the least amount of deaths in Swedish history: https://cornucopia.cornubot.se/2020/10/september-2020-den-mi...

An obvious hypothesis is of course that everyone that "should have died" already died in April and May, at the height of the Covid-19 epidemic.

Regardless, I think the takeaway of the article is correct. There's a month and a half left of the year and even in a worst-case scenario Sweden is unlikely to register statistically significant excess mortality for the year 2020.

At least in Canada and the US, covid death rates are on a sharp upward trend this November, so I don't think this argument holds.

The official mortality data is available from Sweden's statistics agency as a regularly updated Excel file [0], so anyone can do their own analysis and draw their own conclusions. If you're looking for covid related data specifically, the official data for that is available on this arcgis page [1].

[0] https://www.scb.se/en/finding-statistics/statistics-by-subje...

[1] https://experience.arcgis.com/experience/09f821667ce64bf7be6...

> That's higher than 5 of the shown previous 10 years

So in other words, it's basically exactly the median of the last 10 years? That seems to reinforce the title's claim.

But the projection is already for excess deaths over half of the years. (And especially: last year.) That's not fairly-described as "no excess deaths".

And, for the other reasons I listed:

* the projection is highly likely to be an underestimate of the actual 2020 deaths

* an apples-to-apples comparison of the deaths through November 6 in past years would be quite likely to show a notable excess in 2020, given the seasonal nature of the data left out (higher rates of deaths late-November and December) even in normal years

So only a naive interpretation, or tendentious interpretation, of the 2020 year-to-date numbers supports the headline assurance.

So basically we just need to wait until after the years end to know.

What do you think? A month or two into 2021?

https://i.imgur.com/E1M8eMj.png

The fact the prior "spikes" during April-June, which were way smaller than the current November spike (1678 june peak vs 5764 nov peak) didn't already caused a notable increase is mildly interesting. But really Sweden did great overall until recently when the had the first major spike.

Maybe 1000/day extra covid positive cases really isnt enough to move the fatality needle much compared to the average normal deaths... and in many countries recent spikes (Sept-Nov) have also had much lower fatality rates vs positive rates as well.

It's nonsense, you can't just take the average for the year so far and fill it in for December when there are substantially higher deaths in December typically.

Point is, that even assuming winter deaths, Sweden's excess mortality for the year can't at this point get high enough that it'd have been noticed in any normal year. Normality+some winter deaths does not make for a disaster. Yet we see posts here saying it's too early to tell if Sweden is a COVID success or "disaster". That looks like motivated reasoning: under what plausible scenario could Sweden end up a disaster for 2020 with these stats? There'd have to be an actual natural disaster for deaths to become high enough to be noticeable.

Unsure why motivated reasoning in someone else's comments is relevant to this discussion. It's just about facts.

Look up Sweden excess deaths by month rather than analyzing an overall year number that compares up to November with the full year of data for the previous years.

https://ourworldindata.org/excess-mortality-covid

Scaling up to Sweden's population and the excess deaths that they're looking to have will probably be comparable per capita with the 200k from the US so far. Nevermind that Sweden's economy has done worse than its peers that did lockdown, with -5 GDP growth since q4 2019 vs. Denmark -3.9.

So really, no world in which Sweden's approach was successful.

What's your point? That there was a wave of excess deaths? That happens in most countries in Europe, most years, just due to flu and other winter issues. Not every year: Sweden was running 10% lower excess deaths than the 5 year average until the first wave hit, so some of those excess deaths were over-due according to historical trends. That's why it's misleading to point to month-by-month graphs and say for example, look, April is never normally that high, there's a crisis. No, normally a wave much like that one happens a few months earlier. But so what? The exact timing of a normal-sized wave of excess death is no crisis.

the excess deaths that they're looking to have will probably be...

Will probably be normal. That's what this thread is about. There is no evidence of mass death of any kind of Sweden despite that this was supposedly guaranteed without lockdowns. September was one of the least deadly months on record in Sweden. They are on track to experience a basically normal year in terms of deaths.

As for motivated reasoning, it's relevant to this discussion because the thread is full of it. Many of the talking points are absurd. There are people arguing that the Swedes are naturally anti-social and don't meet each other (!), that they're incomparable to other countries due to population density (Stockholm has similar density to London), that they should have locked down because their economy only did a bit better vs other countries (compare against the UK or France and their GDP looks good), and all manner of other things that are either wrong or irrelevant to the reason their death stats are interesting: they didn't lock down and their total, all cause deaths are nothing remarkable. The implication is obvious: lockdowns were pointless, the damage they caused was avoidable, the academic advisors to governments were wrong (except in Sweden), and other countries should unlock ASAP.

> That's higher than 5 of the shown previous 10 years.

Isn't the word for that "median"? i.e. this years deaths are the same as average median deaths in the past 10 years?

Covid didn't become a big thing in Sweden until partway through the year. So the naive scaling which doesn't take that into account will underestimate the problem.

Not to mention that any comparison should be done April 2020 to March 2021 not Jan to Dec.

Why? Do you think that delaying the analysis will make a difference in anyone's conclusions? Do you think COVID will be over in March 2021? If not, why not do a running tally from April 2020 until the disease is eradicated?

Because they're talking about the number of excess deaths in a year. If it's in order to debate a question about Sweden's approach, it should be dated from when Sweden's approach diverged from the approach of others. And it should last 12 months, because that's how long a year is.

Are we now redefining now what a calendar year is just so that we can somehow get to the result we want? that's not science, it's not how it works. You're saying that even though for the past decades we're looking at mortality per a normal calendar year, now we need to redefine this, because the results just don't fit with our opinion and beliefs.

Notably, the link isn't to a paper, it's just to a blind graph of raw data. It's not only bad analysis, it's non-analysis. The only provided interpretation is in the spun headline.

> our 365-day leap-year 2020

It doesn't make much difference to the calculation, but being a leap-year, 2020 has 366 days.

How can something be a disaster if it needs retrospective analysis years later to come to that conclusion?

Lots of public health disasters only become clear in retrospect, sometimes decades or even centuries later: lead-in-gasoline, air-pollution's effects on all-cause mortality & mental acuity, nuclear testing fallout, the collapse of pre-Columbian populations in the Americas.

And lots of media fed panics also only become clear in retrospect, e.g. Y2K.

Like many of these things, disentangling exactly what is what can be hard. If the media had ignored Y2K, I have no doubt that it would have been much worse as many companies, probably on the smaller end, weren't aware random bits of software was susceptible, weren't updated from vendors (who probably had little contact after the sale) about possible problems and updates, etc.

Sure, maybe IBM and Microsoft and the banks and Airlines would all be fine, but if a lot of small to mid-size accountancy firms all of a sudden couldn't process payroll for companies or deal with day-to-day needs of their clients, that could very quickly have gotten really had and had knock-on effects, and that's one industry.

The best case scenario for COVID-19 would have been everyone heeding the media and staying in for a few weeks early on and it petering out. Everyone would have said it was just a media fed panic then, but obviously, it's a bit more complicated in that case, because that didn't happen (or at least didn't happen enough for that outcome).

How the heck was Y2K a “media fed panic”? Media barely knew how to report tech at all, and barely knew how to explain what was at risk with Y2K. They mostly reported it as “might be a problem the clock is counting and we shall wait and see”.

That's not how I remember it, plus what you're saying is inconsistent with how media works generally. Of course they hyped it up, it would have been bad for business not to.

However COVID was described as being like none of those. It was described as "panic right now, exponential growth will overwhelm hospitals within days". If COVID was really a disaster that should have been apparent in Sweden within a week or two of their decision to diverge from the more commonly recommended approach.

In most cases we know history about: the people experiencing disaster and the people retrospectively discovering it have very little social overlap