Infinite Growth (or Decline)

“…the trajectories of all living organisms: rapid, then slower growth, followed by a plateau and eventually some sort of demise.”

Kenneth Boulding, who headed the American Economic Association in 1968, offered a pungent take on this point of view: “Anyone who believes in indefinite growth in anything physical, on a physically finite planet, is either mad or an economist.”

https://www.ft.com/content/db0a7be2-b2d2-11e9-b2c2-1e116952691a

Posted in Data Analytics | Leave a comment

Deaths in Sweden : 2020 compared to 2015-2018

In terms of deaths, was 2020 a remarkable year….? Have a look below, and decide for yourself.

98124 total deaths

What does the graph above tell us….? Well, for one thing, it tells us that most deaths occur among the old, and that the most common age of death is 88 years. Secondly, the graph also tells us that more old people than young people die. But other than that, the graph does not tell us much about how deadly 2020 was. Death counts in terms of absolute numbers is a pretty meaningless metric, whether you want to compare a country’s death toll year-by-year, or comparing different countries. We need more context for making any meaningful comparison, since the number of deaths is heavily influenced by population size as well as by population age structure (among other things).

But first, let’s compare deaths 2020 with the average number of deaths 2015-2018:

deaths 2020 vs 2015-2018

So, what does this graph tell us about deaths 2020…? Well, a couple of things : first, we can see that there were more deaths 2020 vs. baseline for most age groups over 70, while for most age groups under 70, deaths were about the same or even below baseline.

Still, even after this comparison, we still lack enough context to determine whether deaths 2020 were exceptional. Why…? Because population size, in total as well as within each age group is not the same year to year.

So next, let’s look at age group specific mortality, still 2020 vs baseline 2015-2018. For less clutter in the graphs to follow, I’ll use 5 year age bins instead of yearly bins

age group mortality 2020 vs baseline

Above we see a couple of things : first, mortality for ages under 45 is so low that it is not visible in this graph (we’ll fix that in the next graph). Secondly : it’s only when we look at age groups 80 and above that we from this graph can see any major difference between 2020 and the baseline. It’s also important to look at the bars for 2019 : for all age groups where 2020 had mortality above baseline, it sure looks like the year before, 2019, those age groups were well below average. Dry Tinder / Regression to the Mean, perhaps….?

To see mortality for the younger age groups, let’s plot the same graph on a lin/log scale:

age group mortality log scale

Looking very (!) closely at the graph above, we see that for 6 of the 21 age groups, mortality 2020 was above baseline. It’s also interesting to observe that for 18 of the 21 age groups, mortality 2019 was below baseline.

Next, lets look at the growth of mortality per age group, 2019 & 2020 vs baseline 2015-2018:

age group mortality growth 2020 vs baseline

A couple of interesting observations here: first, look at 2019: mortality 2019 was substantially below baseline for all age groups except for 15-19 and 20-24. Should we be alarmed that more 15-19 and 20-24 year olds died 2020…? No. Why…? Because of the “Law of growth of small numbers”. Google it. Or look at the numbers yourself. The important point here is that during 2019, deaths, regardless of age group, were at record lows. Could that fact have any impact on deaths the subsequent year, i.e. 2020…?

Finally, with mortality per age group at hand, we can define “Expected Deaths” as population per age group times baseline mortality for that age group, and with that concept, we can also define “Excess Deaths” as the diff between Actual Deaths and Expected Deaths. Thus, we thereby define “normal mortality” as mortality per age group as it was on average 2015-2018, and derive “Excess Deaths” based on that “normality”:

Expected Deaths

Above graph shows the Expected Deaths for 2019 & 2020 (dashed lines), and the Actual Deaths (x’s). The two Expectation plots look pretty identical up to 70 years, but diverge between 75 – 85, where the expectation for 2020 was higher than for 2019. Notable is that for 2019, the actuals are at or below Expectation, while during 2020, for a handful of the older age groups actuals are higher than expectation. So, clearly, 2020 clearly had some “Excess Deaths” among the old, while 2019 had a “Death Deficit”. Let’s quantify these:

Excess Deaths rel. baseline 2015-2018

Above graph shows “Excess Deaths” 2002-2020, using age group mortality 2015-2018 as baseline.

A problem with the model used above for calculating Excess Deaths is that it does not take the overall declining mortality trend into account – over time, mortality has gradually been declining for quite some time. To address that, the below graph shows what happens to excess deaths 2019 and 2020, when they are obtained by a linear regression based on expected deaths years 2002-2018:

linreg model for excess

The numbers for 2019 and 2020 change a bit, but the “pattern” remains the same – the overall result in terms of outcome is identical : a huge death deficit 2019 followed by moderate excess deaths 2020.

So, was 2020 “A remarkable year” in terms of deaths….? Not nearly as remarkable as 2019.

Posted in Data Analytics | Tagged | 1 Comment

Fler dör av Lockdowns än av Covid

https://www.svt.se/nyheter/vetenskap/miljontals-har-dott-av-restriktionerna

Fattiga betalar priset för de rikas panik.

Posted in Data Analytics | Tagged | Leave a comment

Synthetic mRNA Covid vaccines: A Risk-Benefit Analysis – OffGuardian

Synthetic mRNA Covid vaccines: A Risk-Benefit Analysis – OffGuardian
— Read on off-guardian.org/2021/02/22/synthetic-mrna-covid-vaccines-a-risk-benefit-analysis/

Fancy a flu shot…? You might want to read this first.

Posted in Data Analytics, Epidemics | Tagged | 2 Comments

Covid Sweden – Final mortality based on the now Official 2020 Death Data from SCB

In this short post I’ll present a couple of graphs showing the final, official all cause mortality numbers 2020 for Sweden.

In summary : 2020 was *not*, contrary to popular belief, a very remarkable year in terms of deaths.

Actually, this is pretty much a totally redundant post, since the official numbers released yesterday (2021-02-22) do not change anything from what I presented here , using the SCB preliminary data, already more than a month ago, January 18:th. At that time, the preliminary death toll for 2020 was very close to 98000, and the final number, presented yesterday, ended up at 98124.

Mortality

Sweden Mortality

Mortality 2020 ended up just below that of 2012, exactly as obtained in my previous post of January 18:th using the preliminary data.

Age Group Specific Mortality

Now, with the official numbers available, I have access to yearly age bins also for 2020. Using that data, age group mortality (10 year bins) looks like below:

10 year binned age group mortality

The black dashed line in each plot represents a baseline obtained by the average mortality for each age group taken from the years 2015-2018. Notable is that mortality 2020 is at or below baseline for all age groups below 70, except for age group 10-19 which resides slightly above baseline. The reason for the slightly higher mortality than “normal” for the 10-19 years olds has to do with what I call “The Law of Small Numbers” : 204 children & youngsters in that age group died 2020, and with such a low number, even a very small change year-to-year looks large.

Age Group Specific Mortality Growth vs Baseline

A perhaps better view of 2020 mortality per age group is obtained by looking at the group specific mortality growth factors:

age group mortality growth

As can be seen in the graph above, only 3 age groups (10-19,70-79,90-99) out of 11 had mortality significantly above Baseline. And as mentioned above, the higher than “normal” mortality for the 10-19 year olds is explained by the Law of Small Numbers, meaning that all “Excess Deaths” (whichever way we’ve chosen to define “Excess” – see here for info on how ambiguous the notion of “Excess Deaths” is ) occurred in age groups 70-79 and 90-99.

Posted in Data Analytics | Tagged , , | 3 Comments

An analysis showing that much of Covid-“science” is far from SCIENCE

Posted in Data Analytics | Leave a comment

How deadly is life itself… and how deadly is Covid…?

How dangerous is life itself ? What’s your risk of dying – regardless of cause – during 2021…? And what’s your risk of dying of Covid…?

Well, in fact I don’t know the exact probability for *you* specifically to die of any cause, or of Covid specifically, but I can provide an estimate for the “Average Person”.

In fact, in what follows, we’ll define several Average Persons, one per age group, and with 10 age bins, we’ll end up with 10 Average Joe’s. So, if you’d like to know your risk of dying, on average, pretend you are the Average Joe for your age group. All numbers based on Swedish stats.

Let’s start with the risk of dying 2021 of any of the myriad of viruses/diseases/accidents etc that might kill you:

The below graph shows the probability/risk of dying per age group, on average, derived from years 2015-2018:

probability of dying

If you happen to be younger than 60, then your risk of dying this year is barely visible on the graph above, the plot is almost flat at zero until you – like me – passed 60 years of age. The actual probability of a person between 50-59 years of age to die is (on average) 0.003, that is, 0.3% or 3 out of 1000. For the younger age groups, the risk is significantly smaller.

Even if you happen to be between 70-79 years old, your risk of dying is not very high, about 0.02 or 2%. Then, after having reached 80 years, the risk starts to accelerate, and when you are above 90, you face a significant risk (27% on average) of dying this year.

All the numbers above have nothing to do with Covid – the numbers are from 2015-2018, i.e. long before anyone had heard about the “Pandemic”.

The main lesson from the above is that young(er) people face a very limited risk of dying, whereas the longer you’ve managed to survive, the more certain is your upcoming death.

Next, let’s see what the probabilities for Covid specific events are, still using the same age bins – we’ll look at probabilities for getting a positive test, getting hospitalized at ICU, and finally dying of Covid:

Covid impact per age group

There are 5 plots on the graph above. Still the same age bins from above. The first (blue) one shows the probability of getting a positive test per age group, the next the probability of ending up in an ICU, and the third shows probability of dying from Covid.

Let’s focus on the third, red plot, the probability/risk of dying from Covid. Notice something interesting…? The “form” of the plot is very similar to that of the all cause death graph we just looked at above – the key point to observe is that if you are younger than (say) 80, your risk of dying of Covid is very small, and the younger you are, the lower the risk. How low…? Let’s take the same age bins as above:For age group 50-59, the Covid risk of death is 0.0002, for 70-79 year olds the risk is 0.002, and for 90+, the risk is 0.04.

You can compare the “form” of the plot showing the risk of dying of Covid (the red plot) with the form of the all cause death risk plot (black) (which is the same plot as in the first graph) – the impact of Covid in terms of deaths per age group is very similar to the impact of all cause deaths for the age groups: basically, what it means is that people who die from Covid are to a very large extent the same cohort that “normally” die, but from other causes.

Looking at the scales, we can see that for all age groups, the risk of dying of Covid vs the risk of dying of any of the multitude of other mishaps that could get you killed during 2021 is about an order of magnitude smaller.

For the younger age groups, the risk ratio is even smaller.

To understand what all this means, it might be helpful to consider a hypothetical scenario where Covid, instead of killing the old and frail, would have killed primarily the young & healthy, let’s say those below 50, or even worse: children – that nightmare scenario did thankfully not materialize. But if that scenario indeed had materialized, the red vs the black curves above would have had very different forms.

Another illustration of the similarity of Covid and non-Covid deaths, in terms of which age groups are mainly targeted, is given by the below graph. The blue shows the distribution (still 10 year age bins, that’s why the distributions are “contracting” between the bins) of all cause deaths, the orange shows the distribution of Covid deaths:

age death distribution Covid vs All Cause Deaths

For the top 3 age groups, the distributions are almost identical. Then, for the younger age groups, the impact of Covid diminishes rapidly.

Finally, let’s look at a timeline of “Excess Deaths” for Sweden, 2015-2020:

Excess deaths

A couple of things worth to notice: first, look at 2019. Second, look at 2020. Any reflections…?

Posted in Data Analytics | Tagged | Leave a comment

Seasonality of Deaths (at relatively high Latitudes)

Playing around with the SCB death data, I just discovered something that is perhaps bleedingly obvious (or not), but anyhow potentially interesting : Seasonality of deaths up here in the far north:

More people die during early & late year than during the rest of year.

First, let’s look at monthly mortality during the period 2015 – 2020 for 4 different age groups:

monthly deaths 2015 – 2020

The grey shadow shows the mortality range based on 2015-2018 data, and the red plots show the actuals. What’s interesting is that all age groups except the youngest (0-64) have a clear seasonality pattern, both in range as well as in actuals. Or maybe it’s more correct to say that it’s interesting that the 0-64 age group does *not* exhibit a clear pattern of seasonality… Any guesses on why that is so…?

Perhaps “darkness and cold kills…. but mostly the old…” ???

Next, let’s look at the distribution of deaths over day of year for the four age groups:

distribution of deaths on day of year

The above graph has 4 “ViolinPlots”, each showing the distribution of deaths over day of year for an age group. You can imagine tilting each Violin 90 degrees and eliminating the bottom half to better see that these are indeed distributions.

Anyhow, looking carefully at the Violins, we can see that the bottom part (early days of year) is slightly “fatter” than the midpart (summer), and that the top part (end of year) again is fatter than the middle. The difference is more notable in the older age groups than the 0-64 group, that is, older people’s deaths seem to be more impacted by seasonal variation than younger people’s.

Posted in Data Analytics | Tagged | Leave a comment

Final Report on Swedish Mortality 2020, Anno Covidius

Fourier Cycles & Epicycles

[UPDATE 2021-02-23 : SCB, the Swedish Bureau of Statistics, released their official 2020 death data yesterday. As I predicted, the official data fully concur with the preliminary data of January 18:th that I used in this post. For more details on the official numbers, see this recent post]

2020 was an “interesting” year

2020 – “The Year of Doomsday Modeling”

Scientists and researchers from lots of different disciplines produced lots of theories and mathematical models intended to explain and predict Covid and its impact.

Models that in most cases eventually proved to vastly inflate and overestimate the risks & resulting deaths. It appears that most of these scientists fell in love with their own models & theories, and thus failed to take real world data and reality into account.

Recall all the doomsday models predicting massive deaths already during the “First Wave”…?

E.g. the one by Professor Ferguson of Imperial College, or this one from Belgium, predicting, in line with the predictions of Ferguson’s model, a median of 96000 deaths in Sweden by May 2020 (!), unless strict Lockdowns were immediately put in place….?

2020 was also the year when media declared Sweden as a failed state, the “World’s Cautionary Tale”, predicting massive deaths unless Sweden, like almost everybody else, fully LockedDown.

And 2020 was the year when a lot of economies, public as well as private, businesses and entire industries went bust, because the response governments took to mitigate the pandemic.

During 2020 it also became obvious that True Leadership, of the kind demonstrated e.g. by Churchill during WW2, has pretty much vanished from politics, media as well as from society in general – amplifying fear and panic is the last thing you should do as a Leader during a crisis.

2020 was also the year when China smashed the western world so hard that it will take decades for the west to recover, if indeed ever.

And finally, 2020 was the year when we stopped living due to fear of dying.

So, how deadly did the Pandemic turn out to be ? How many “Excess Deaths” was there…? How proportional were the responses of the various governments ?

The Swedish Outcome

Now, mid January 2021, we have the Swedish result (still preliminary, but perfectly good for overall conclusions): about 98000 deaths for *the full year* – not 96000 by May 2020 (!) , as the models mentioned above (and many others) predicted.

Still, 98000 yearly deaths is clearly more than the normal ~92000 (2019 with it’s 88766 deaths was a true outlier!) of the past few years, so clearly something has been going on 2020. Let’s see if we can figure out what’s been going on, and how much of an outlier 2020 really was…?

Can we trust the preliminary data…?

But first, let’s briefly discuss the realiability of SCB’s preliminary death data: I’ve got a huge number of comments over the past year that :

“you can’t trust the prel. data, there is a huge lag, the numbers will increase a lot by subsequent reporting updates, just wait two more weeks and you’ll see…!”

I’m not sure on what these commentators base their opinions on, but mine are based on the data, as well as on what SCB themselves state about the data lag and it’s expected length and the effect of later updates: SCB.se:

“Statistik för två veckor tillbaka i tiden väntas inte förändras i någon större utsträckning. “

For the non-Swedish speakers, I’m sure Google Translate can help out in making sense of the above quote.

To see that the statement by SCB holds, let’s look at the number of monthly deaths per reporting date, and see how the data changes from one report to next, and then stabilizes, goes flat, when 3 weeks have passed since the first report after month end for that month:

look at the graph below. Take any plot representing a single month e.g. the red plot representing April, and observe how the plot goes basically flat on the third reporting occasion after the month has ended. So, data 3 or more weeks back do not change substantially, exactly as SCB state in their press release above.

If we for each month calculate the difference between deaths the latest reporting date (Jan 18:th), and deaths the third weekly report after end of each individual month, the average monthly update (discounting December, which is most likely have a larger update somewhat later than the “normal” 3 weeks) is 75 deaths.

Now, since December & January are holiday months with both Xmas & New Year, the data reporting lag is somewhat larger for December data than for the other months, but for all practical purposes, for an overall result on whether 2020 was the contemporary equivalent to the Spanish Flu, it doesn’t really matter if the December deaths eventually, when the official 2020 numbers are released, are going to be 100 or even 1000 larger than what was published today.

That possible increase will not change the overall 2020 outcome in any significant way. I’m not after Rocket Science Precision to the n:th decimal in these analyses, and more importantly there are way more significant sources of massive uncertainty in any calculations on Swedish mortality than the possible data lag in SCB’s prel. death data, see below.

Reporting lag

Another, perhaps more significant source of uncertainty regarding the numbers on Swedish mortality stems from political decisions over the past couple of decades, and is described in this (Swedish) article from Göteborgsposten. The basic message is that the true Swedish population size is unknown (!) and could very well be several hundred thousand larger than the official numbers say.

2020 – Highest Number of Deaths since 1918 !

The above headline is typical of most media – alarmistic & fear mongering. And it is true – Technically true, but semantically very misleading. Before going into details about why that headline is misleading, let’s first look at some features of the chart below, showing the monthly number of absolute deaths 1900 – 2020:

Absolute deaths

There’s several interesting anomalies (‘outliers’) in this data, the most obvious one is 1918, the year of Spanish Flu. But also interesting is 1942/43 – in the middle of WW2 deaths *dropped* substantially…! Maybe since almost all Swedish males were drafted, and thus got well fed and cared for by the military, with strict rules of behavior,lots of fresh air and physical activity, and few opportunities for boozing, partying and other types of mischief..?

Here’s a link to an old SCB document on causes of death 1942.

Other notable things on the chart above is that the overall death trend was going down until 1960’s, after which the trend changed direction, the number of deaths started growing, reaching a plateau in early 1980:ies. Any guesses why…?

While talking about outliers, take a look at 2019 in the colorful graph above – deaths 2019 dropped significantly from recent earlier years. We’ll return to that fact below. And finally, 2020: Yes, it it definitely an outlier in terms of absolute deaths – we will have to go all the way back to 1918 for a year with higher death toll.

Demographics – Population Size Matters

Swedish population grows year by year. The more people, the more deaths. Pretty obvious to all but media, it appears. Therefore, comparing absolute deaths instead of deaths per capita, is simply misleading (accidentally or intentionally).

Unfortunately, it’s not just media that enjoys this kind of “Technically true but Semantically misleading” fear mongering and alarmism – even the Swedish Government Bureau of Statistics, SCB, have several times over the course of the pandemic, e.g. here, released news bulletins where they compare different years in terms of absolute deaths. Simply amazing!

But here, in this blog post, we prefer to present semantically meaningful data; here we care for facts & numbers that can help us understand reality.

So let’s adjust the Swedish death numbers by population – as you can see below, population is continuously growing, and growth is accelerating over the past decades.

Population Adjusted Deaths (Mortality)

Below mortality (deaths / capita) 2002 – 2020.

Mortality 2020 ended up on par with that of 2012.

mortality

Let’s next look at mortality trends over a longer period of time. Below chart presents deaths per capita (“mortality”), based on the same death data as was used a bit earlier in this post. Can you spot the Pandemic…?

Can you spot the Pandemic….?

Looking at the full dataset, we can clearly see that there’s a declining overall mortality trend, very likely caused by advances/improvements in population health, medical technology and care,nutrition and other socioeconomic factors.

However, that trend is not monotonic, instead there are clearly visible fluctuations in the data as well as in the trend, e.g. the period of increasing mortality starting around 1960, followed by another declining trend from early 1990 onwards. That latter trend might, or might not, now from 2020 onwards, be broken, e.g. by changes in demographics, as explained in this post from SCB:

“Under 40-talet föddes det väldigt många barn i Sverige vilket gör att vi kan vänta oss att antal döda per år kommer att öka de närmsta tjugo åren.”

Or perhaps we’ll see that 2020 indeed was a real outlier due to the pandemic, and the coming next years will continue the downward mortality trend.

Below a trendline fitted to the full dataset, among other things showing that some caution is always in order in reading trend lines – Math’s & Stat’s both are very unforgiving Gollums in answering your questions – what you ask for is exactly what you get… be sure to ask the right question…. :

Misleading trendline

According to the trend line above, it looks like the entire period from 1980 onwards has had “excess mortality”, while the period between 1920-1960 has significant mortality deficit – two different periods of “outliers”…

For more info on why care must be taken whenever trying to understand trendlines & various baselines, look no further than here.

For a slightly more relevant trendline, let’s change the start of the dataset to 2010:

somewhat more meningful trendline

Above we should notice that if we consider 2020 as an outlier, then so should we also consider 2019 as an outlier – mortality 2019 was exceptionally low, and has clearly contributed to the high mortality of 2020 – “Dry Tinder”.

More Demographics – Age Matters

The older the population, the more deaths. Old people tend to die more frequently than young people, in general.

Below graph shows the Y2Y growth factors for different age groups compared to year 1900. Note that the y-scale is not percent, but the actual growth factor.

Below, the proportions of Swedish population by age group and year:

Age Group Proportions

We can also check the age group growth with other baseline years, below six different baselines. It’s clear that the proportion of old people is growing.

So, with a population growing older year-by-year, we need to adjust for age in our mortality calculations to obtain fair mortality numbers to compare. Let’s do so:

Age Adj. Mortality

Taking population age structure (proportion of old vs young) into account, we arrive at an age adjusted mortality 2020 on par with that of 2013.

We can also look at the same numbers in terms of “Life Expectancy”:

Life Expectancy

Notice above that 2019 really stands out.

Finally for this section on the importance of population age structure, let’s take a look at the age group specific mortality trends, and comparing them to two different baselines, 2015-2018 and 2015-2019, respectively:

Age group mortality trends

Interesting to notice that for two of the age groups above, the youngest and the oldest, mortality 2020 was below both baselines, while the other two were above.

Even more Demographics – foreign born vs native born

About 1 in 5 people living in Sweden 2019 were foreign born.

There are significant differences in the impact of Covid between native vs foreign born in Sweden. Let’s first look at weekly mortality for both groups:

weekly mortality, native vs foreign born

As we can see from the graph above, from the red and green dashed lines, foreign born typically have much lower mortality rate than native born. The reason most likely being that the average foreign born person in Sweden since a decade or two ago is young, mostly male, i.e. a male migrant/refugee, in his 20-30:ies, and this of course has an impact on mortality of that cohort – young people die less frequently than old people.

Another thing to notice from the graph is that Covid impact hit the foreign born cohort harder, relatively speaking, than the native. How much harder…? Let’s look at the respective weekly mortality growth rates:

native vs foreign mortality growth vs baseline

Above, we can see that foreign born were hit by Covid substantially harder than native born during the first spring wave: foreign born mortality hit almost factor two of the baseline, while corresponding number for native born – and remember, native born are on average much older – was below 1.4. Also during the Second Wave Nov-Dec, foreign born were hit harder by Covid than native born.

What can explain this striking difference in mortality, where a significantly younger cohort (the foreign born) got hit much harder by the virus than the substantially older and supposedly more frail cohort of native born…? I don’t know, but a reasonable guess could be socioeconomic factors: it’s likely that vastly larger proportion of the native born have much better socioeconomic standing, better jobs, better living conditions, possibility to work from home, travel by own car etc, while the foreign born to higher extent have jobs where physical presence is required. So, my guess is that Zoom is much more frequently the tool of the trade for native born than for the foreign born, and public transport is a much more frequent mean of commute for foreign born than for native born during the pandemic.

Pandemic accounting based on the Gregorian Calender – Regression to the mean

I’m not sure viruses in general care much about calendar’s and years. That is, there’s really no obvious reason for us to measure the impact of a virus based on calendar years.

Calender is a human invention, used by humans to keep track of the periodic patterns of days slowly passing by, governed by our rotation around the sun, and the moon rotating around us.

Let’s next look at what happens if we instead of measuring deaths per year, measure deaths by pairwise years: Let’s look at mortality obtained by looking at pairwise consecutive years:

Mortality pairwise years

Now, when combining deaths 2019 – a year with the lowest mortality ever – with deaths 2020, we see that average mortality for 2019/2020 is very much normal. The large number of deaths 2020 is offset by the low number of deaths 2019. Regression to the Mean in action.

Of course, there’s nothing stopping us looking at deaths pairwise years. We can of course combine deaths from more than two consecutive years and look at the mortality obtained:

As mentioned above, I don’t believe viruses care much about calendar years, so let’s also look at seasonal deaths. Below, I’ve defined “season” as going from October 1:st to September last. Unfortunately we’ll have to wait until October 2021 to see the 2020/2021 season results:

seasonal deaths

Covid Impact – young vs old

So now, let’s stop beating the mortality numbers any further, and instead look at the impact of Covid

Covid impact per age group

Above, we can see that while most age groups (except the really young) get Covid (or at least a positive test), the more severe consequences occur among the elderly: in particular, if you are under 70 years of age, your risk of dying from Covid is very low. Also, if you are 30 or younger, your risk of needing ICU care is minimal.

So Covid is a virus that primarily is dangerous for the elderly, not for the young.

The graph below shows mean, median and mode for Swedish deaths 1968 – 2019 (the Swedish 2020 death data binned on 1 year intervals are not available yet), and the text box upper left shows the corresponding stat’s for Covid. Notable is e.g. that median age of Covid death is higher than median age of all deaths 2019.

mean, median,mode of death

So, Covid is dangerous for the old, not for the young. On average. Still, the young got hit very hard by Covid – not by the virus itself, for the vast majority of the young the Covid was no worse than a traditional flu – but by the responses the governments took to mitigate the spread of the virus : schools closed, career opportunities that vanished, a year or more of lost life earnings, sports activities terminated or restricted, restaurants forced to close or limit their opening hours,social interaction reduced, etc etc. It remains to be seen over the coming years what impact these unprecedented limitations on the lives of all these young people worldwide will have.

Swedish Crisis & Contingency Planning and Management – or the lack of it ?

“Our hostpitals are at risk of becoming overwhelmed”

Yes, definitely – look at this graph showing ICU bed usage over the past 6 years:

The spring Covid wave is clearly visible, with about 2x the normal number of beds needed, as is the Nov-Dec second wave, with 40% excess demand. BUT….:

Let’s have a look at the Swedish hospital capacity, in relation to the rest of the EU:

Only Lichtenstein has lower hospital capacity per capita than Sweden, at least 2018, and I doubt Swedish position has improved since then.

Below the trend 2007 – 2018 for Swedish hospital capacity:

Next, let’s look at the evolution of Swedish ICU beds, comparing 1993 to 2018:

ICU-capacity has gone down by a factor of 8 (!) in 25 years. So whatever emergency or crisis that would have hit Sweden, the capacity to deal with that crisis was (and is) simply not there.

No wonder the hospitals are at risk of being overwhelmed!

As a side note, here’s another example of these severe failings in crisis & contingency planning is the forrest fires of 2018, when the national fire fighting capacity was way below bar to deal with the fires. It all ended – eventually, when massive areas of forrests already had burned down, and several small towns/villages were at risk of burning down – with hundreds of Polish fire fighter coming over from Poland to clear up the mess caused by years of political mismanagement.

For more info on the forest fire’s of 2018, here’s what the Government writes in a post mortem analysis (Swedish).

The bottom line of all this is that Swedish governments since decades back have taken decision upon decision that have contributed to make the Swedish society very fragile, incapable of managing any unforeseen crisis or disaster. Let’s pray no war will break out in Sweden – if it does, we are doomed.

The high Covid death toll among our elderly, which clearly is due to decades of insufficient investments in Swedish medical care, and care of our elderly and frail, is a clear demonstration of this.

Conclusion

So, what do we make of all the above facts & figures ? Did Sweden really experience a severe deadly pandemic, a “once in 100 years flu”, with people dying in unprecedented numbers during 2020…?

A pandemic of a kind we’ve not encountered since the Spanish flu 1918…? Or did Sweden experience a severe flu with mortality about the same as past severe flu’s that tend to occur once or twice in every 20-30 years or so…?

I don’t particularly like the notion of “Excess Deaths”, since most of the time it’s impossible to know the answer to “Excess to what, exactly?”, as explained here, but below graph shows six different values for Swedish Excess Deaths 2020. They range from 7000to 1900, depending on the chosen calculation method and baseline.

Pick the number most appealing to you and your purposes.

“Excess Deaths”

In textual form, the Swedish “Excess Deaths” for 2020, computed by comparing “Expectation” vs outcome, with two different baselines, are then as follows:

2020 EXCESS DEATHS : 
Absolute excess deaths cmp baseline 15-18 : 6429 
Absolute excess deaths cmp baseline 15-19 : 6978 
Population adjusted excess deaths cmp baseline 15-18 : 3251 Population adjusted excess deaths cmp baseline 15-19 : 4334 
Age adjusted excess deaths cmp baseline 15-18 : 1901 
Age adjusted excess deaths cmp baseline 15-19 : 3146

Applying the same “Excess” computations to 2019 instead of 2020, instead of “Excess Deaths” we get a significant “Death Deficit”, regardless which calculation method or baseline we use:

2019 DEATH DEFICIT :

Absolute death deficit cmp baseline 15-18 : -2745

Absolute death deficit cmp baseline 15-19 : -2196

Population adjusted death deficit cmp baseline 15-18 : -5384

Population adjusted death deficit cmp baseline 15-19 : -4307

Age adjusted death deficit cmp baseline 15-18 : -6146

Age adjusted death deficit cmp baseline 15-19 : -4916

So even if Covid never would have arrived at all during 2020, the death toll of 2020 would almost certainly been significantly higher than “normal” anyway, due to Dry Tinder-effects.

You’ll be the judge over whether the numbers above justify the “World’s Cautionary Tale” designation, or the restrictions on freedom, liberty, future and normal life that have been put in place.

My personal take on Covid 2020 in Sweden is as follows:

  • Yes, Covid 2020 was real (and continues to be real at least until spring 2021, as all seasonal viruses). The number of deaths 2020 was higher than it should have been, which ever way we define “Excess”. Not exceptionally higher, and far from all the disaster scenarios painted by media, politicians and failed scientists.
  • Was Covid 2020 our generation’s “Spanish Flu” ? No. Far from it, as can be seen in the graph showing 1918 above, and by comparing mortality rates, where non-age-adjusted mortality 2020 is on par with that of 2012, and age adjusted mortality 2020 on par with 2013.
  • Was the Swedish Government’s response adequate ? To a large extent yes. Until they panicked and lost their mind in November 2020, and introduced “The Swedish Enabling Act“, a form of legislation that is a disgrace to any nation pretending to be democratic.
  • Where “The Strategy” failed was in protecting the frail and elderly, particularly in the care homes. The strategy also failed in overall crisis & contingency planning & management, where various governments since the early 90:ies have radically reduced investments and capacity in health care, care of elderly as well as many other vital parts of the societal safety net. So, the frequently repeated “Isolate, or our hospitals will be overwhelmed!” mantra was primarily caused by several decades of catastrophic political decisions and priorities regarding medical care and other critical societal function investments and resources, as much as by the virus itself.
  • What the future brings will be seen by those who survive. Myself, I’m afraid that more doom & gloom will follow for a long time in the tracks of the “2020 Covid Experience”, even if we should manage to eliminate the virus, e.g. by vaccine, during 2021. The psychological effect on populations having spent a year or more in Lockdown, thus missing most of what makes life and living worthwhile, will be interesting to observe, as will be whether social interaction patterns and behaviors eventually return to normal, or whether our future social interactions will be so deeply ingrained by Anno Covidis that we will, similar to Pavlov’s dogs, continue regarding fellow human beings as potentially deadly virus vectors.
  • Similarly, as this recent article (Swedish) shows – 90000 (!) medical treatments cancelled during 2020 – we will also have to expect further “Excess Deaths” down the road, where these deaths are only indirectly caused by Covid.

This post will end my own way too long and costly focus on Covid. I will now return to topics that are more rewarding and pleasant to work on.

[A final note: in case a significant number of my readers are interested in trying to reproduce the main numbers of my analysis (nr of deaths, mortality,age adj mortality, excess deaths etc), let me know in the comments and I’ll consider providing a link to where the Python (3.8.6) and Pandas (1.1.0) script I’ve used to analyse the most important mortality data can be downloaded.]

[UPDATE : Apparently someone has made a translation of this post to Italian –

https://www.miglioverde.eu/inchiesta-svezia-la-mortalita-del-2020-e-la-stessa-di-quella-del-2012/?fbclid=IwAR3hAWva741YpRKCuOO34EhpFR0qu56ZVP-4b0X02FAS-trQkKhk3F35t9M ]

[UPDATE 2021-01-26 : While working further on this data, I discovered an anomaly regarding age group mortalities for years before 2020. They didn’t match when I compared mortalities derived from the official data vs those derived from the prel. data.

After lot’s of debugging I found the source of the difference, but didn’t understand what caused it, but after having been in touch with SCB, who were fast answering and very helpful, I learned that they for the age based prel. data, count age of death as age reached when death occurs, while the official death data I’ve used for the age group mortality rate calculations used age reached end of year.

This difference in how age is recorded resulted in the age of death from the prel.death data in some cases (those where people died before their birthday) being a year lower than the records I used, which in turn “shifts” some deaths from older – particularly the 90+ group – to younger groups. Which in turn impacted the mortality baselines I use.

In practical terms, this difference has minimal impact on the numbers (and none on the overall results/conclusions of this post), but since I put quite a lot of time & effort into “debugging” the anomaly, before finally – with the help of SCB – understanding what was going on, I thought I’d provide an update here.

Below is the updated age grp mortality chart, where the main difference cmp the previous definition of age at death is that the 90+ group now is above baseline, with previous definition of age at death it was below. As a consequence of this “shift” of age of death, the younger groups have now lower mortality cmp baseline than before.

Age Group Mortality

And here’s the summary of “Excess Deaths”, calculated using the changed definition of “age at death”, also using the latest SCB weekly report death numbers from 2021-01-25:

2020 EXCESS DEATHS :

Absolute excess deaths cmp baseline 15-18 : 6569

Absolute excess deaths cmp baseline 15-19 : 7118

Population adjusted excess deaths cmp baseline 15-18 : 3391

Population adjusted excess deaths cmp baseline 15-19 : 4474

Age adjusted excess deaths cmp baseline 15-18 : 1996

Age adjusted excess deaths cmp baseline 15-19 : 3248

Finally, below an “Excess Deaths” timeline 2015-2020, calculated with mortality 2015-2018 as baseline. Noteworthy in particular is 2019.

excess deaths timeline
Posted in Data Analytics | Tagged | 39 Comments

Sweden Mortality Update 2021-01-11

can be found here

Posted in Data Analytics | Tagged | Leave a comment