6 ms·

Somewhat deceptive submission title ("No Excess Deaths in Sweden Compared to the Last the Years") and presentation, which might lead the inattentive to think 20

by gojomo 6y ago

Somewhat deceptive submission title ("No Excess Deaths in Sweden Compared to the Last the Years") and presentation, which might lead the inattentive to think 2020 is on a pace for fewer total deaths in Sweden than all or most previous years. Chart declares itself to be data "as of November 6". That's the 311th day of our 365-day leap-year 2020. Thus naively scaling the 77,300 number to the full year would predict ((365/311) * 77300 =) 90,721 deaths in 2020. That's higher than 5 of the shown previous 10 years. But also, the deaths as of November 6 may not be a complete total: often such data series continue to be updated (increased) for a few weeks after the calendar date. 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. (And that's not even counting the Covid-19 case surge happening now in Sweden as with most of Europe.) Everyone trying to draw final conclusions from Sweden is making a call too soon based on incomplete data. We probably won't truly understand whether Sweden is on-net a Covid-19 success or disaster until retrospective analyses balancing all factors in late 2022 or 2023.

marioletto 6y ago
Not to mention that any comparison should be done April 2020 to March 2021 not Jan to Dec.
- nickff 6y ago
  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?
  - pessimizer 6y ago
    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.
    - mns 6y ago
      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.
josephcsible 6y ago
> 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.
- gojomo 6y ago
  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.
  - dmix 6y ago
    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 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.
- whimsicalism 6y ago
  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.
  - thu2111 6y ago
    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.
nxmnxm99 6y ago
How can something be a disaster if it needs retrospective analysis years later to come to that conclusion?
- gojomo 6y ago
  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.
  - mdifrgechd 6y ago
    And lots of media fed panics also only become clear in retrospect, e.g. Y2K.
    - kbenson 6y ago
      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).
    - eyelidlessness 6y ago
      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”.
svrb 6y ago
> That's higher than 5 of the shown previous 10 years. That's a shockingly dishonest way to make a claim which could just as well have been phrased "that's lower than 5 of the previous 10 years".
- iannus_chev 6y ago
  Shockingly dishonest?? Why are you so quick to deliberately misrepresent a single line in the comment, which 1) isn't his argument and 2) he qualifies? The post title is a wild claim and should be treated as suspect.
  - thu2111 6y ago
    The post title is not a wild claim, it's a factual claim based on the data so far. Of course Sweden may well have more than the 1000 excess deaths/week needed to catch up with a normal year, we'll have to wait and see. But as claims go it's not really possible to be less wild: it only seems that way in the context of the extreme predictions made earlier in the year. Which it's apparent now were wrong.
sorenjan 6y ago
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-subject-area/population/population-composition/population-statistics/pong/tables-and-graphs/preliminary-statistics-on-deaths/ https://www.scb.se/en/finding-statistics/statistics-by-subje... [1] https://experience.arcgis.com/experience/09f821667ce64bf7be6f9f87457ed9aa https://experience.arcgis.com/experience/09f821667ce64bf7be6...
randyrand 6y ago
> 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?
- FartyMcFarter 6y ago
  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.
dane-pgp 6y ago
> our 365-day leap-year 2020 It doesn't make much difference to the calculation, but being a leap-year, 2020 has 366 days.
EdwardDiego 6y ago
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-multiple-organs-recorded-in-long-covid-cases https://www.theguardian.com/world/2020/nov/15/damage-to-mult...
- onecommentman 6y ago
  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 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 https://en.wikipedia.org/wiki/Quality-adjusted_life_year
- absolutelyrad 6y ago
  One of the major impacts of CVOID is likely going to be on men's fertility.
bjourne 6y ago
> 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-minst-dodliga.html 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.
- sudosysgen 6y ago
  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.
newacct583 6y ago
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.

by gojomo 6y ago

marioletto 6y ago

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

nickff 6y ago
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?
- pessimizer 6y ago
  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.
  - mns 6y ago
    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.

josephcsible 6y ago

> 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.

gojomo 6y ago
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.
- dmix 6y ago
  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 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.
whimsicalism 6y ago
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.
- thu2111 6y ago
  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.

nxmnxm99 6y ago

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

gojomo 6y ago
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.
- mdifrgechd 6y ago
  And lots of media fed panics also only become clear in retrospect, e.g. Y2K.
  - kbenson 6y ago
    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).
  - eyelidlessness 6y ago
    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”.

svrb 6y ago

> That's higher than 5 of the shown previous 10 years. That's a shockingly dishonest way to make a claim which could just as well have been phrased "that's lower than 5 of the previous 10 years".

iannus_chev 6y ago
Shockingly dishonest?? Why are you so quick to deliberately misrepresent a single line in the comment, which 1) isn't his argument and 2) he qualifies? The post title is a wild claim and should be treated as suspect.
- thu2111 6y ago
  The post title is not a wild claim, it's a factual claim based on the data so far. Of course Sweden may well have more than the 1000 excess deaths/week needed to catch up with a normal year, we'll have to wait and see. But as claims go it's not really possible to be less wild: it only seems that way in the context of the extreme predictions made earlier in the year. Which it's apparent now were wrong.

sorenjan 6y ago

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-subject-area/population/population-composition/population-statistics/pong/tables-and-graphs/preliminary-statistics-on-deaths/ https://www.scb.se/en/finding-statistics/statistics-by-subje... [1] https://experience.arcgis.com/experience/09f821667ce64bf7be6f9f87457ed9aa https://experience.arcgis.com/experience/09f821667ce64bf7be6...

randyrand 6y ago

> 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?

FartyMcFarter 6y ago
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.

dane-pgp 6y ago

> our 365-day leap-year 2020 It doesn't make much difference to the calculation, but being a leap-year, 2020 has 366 days.

EdwardDiego 6y ago

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-multiple-organs-recorded-in-long-covid-cases https://www.theguardian.com/world/2020/nov/15/damage-to-mult...

onecommentman 6y ago
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 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 https://en.wikipedia.org/wiki/Quality-adjusted_life_year
absolutelyrad 6y ago
One of the major impacts of CVOID is likely going to be on men's fertility.

bjourne 6y ago

> 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-minst-dodliga.html 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.

sudosysgen 6y ago
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.

newacct583 6y ago

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.