Canadian 2020 Mortality Structure Dynamics - v3.0

Main Findings Graph Explanation East West Atlantic Final Thoughts

Statistics Canada's data on excess mortality in 2020 is regularly updated and available for public to study. Here, we'll try to:

This is the third iteration of analysis. It improves gradually on healthy criticism and feedback, on new data sources and on corrections that critics bring in. Your feedback is always welcome!

Data Sources

Our major data source will be the following two tables:

Statistics Canada. Table 13-10-0792-01 
Adjusted number of deaths, expected number of deaths and estimates of excess mortality, by week, age group and sex
https://www150.statcan.gc.ca/t1/tbl1/en/tv.action?pid=1310076801

Coronavirus disease 2019 (COVID-19): Epidemiology update
https://health-infobase.canada.ca/COVID-19/epidemiological-summary-covid-19-cases.html

Past iterations used a different data table to predict expected 2020 mortality from past years data. Different approximation methods were attempted. First iteration turned out a highly optimistic baseline estimate, resulting in overestimating excess mortality. Second iteration closely re-implemented the official baseline estimate and verified the validity of the official statistics assumptions. This third iteration parts with raw yearly data and uses the now verified expected baseline and excess mortality data from the official table.

Back to Top ↑

Main Findings

Age distribution
Age Graph

So, we can expect that excess mortality will follow similar pattern. Or can we?

Excess Mortality %
Date ranges differ from province to province, depending on data availability. For more details, see "Data ranges for provinces and geographical regions" section.

WTH? Pretty much every Canadian province shows the reverse trend! Younger ages experienced the highest percentage increase, older people - significantly less!

Those are percents, of course - absolute numbers should look less scary. Vast majority of all deaths are old people, not young, after all. So, even a very large percentage increase in the "young" group applies to a very small total numbers. And vise versa - even a tiny percentage increase for older people should produce much larger totals. We saw exactly that in the first iteration. Now that we know that first iteration was overestimating excess deaths in older population, we've adjusted the baselines - but the general expectation should still hold. Does it?

Excess Mortality #
Date ranges differ from province to province, depending on data availability. For more details, see "Data ranges for provinces and geographical regions" section.

Ugh! For some reason, only Quebec looks "as expected" - with reasonably smooth spread across age groups, closely following our first logarithmic image. In the other provinces, large youth mortality increase remains high even in the absolute numbers!

Youth Mortality

That's not as strange as it may seem. Yes, since the first iteration, we've got more refined data and more accurate methodology. But this methodology improved accuracy only for older ages. For younger ages, our very first methodology was already as good as the new one. This is why our major finding - the unproportional youth mortality - does not look any better. In fact, with decreased share of elderly mortality, it now looks worse!

Another thing not changing over time - the ignorance. We still don't know what kills the young people. No data on causes of deaths, so we can only guess... and look asome other regional stats. Drug overdoses? Part of it for sure! (Calgary Herald: Opioid Deaths in ALberta) Depression and suicide? Does not look like it... at least not yet. (CTV News: Suicides Trends Down) Gang violence? Accidents from new trades and hobbies instead of restricted ones? Road accidents? Not COVID: young people are the least affected by virus directly, with each such fatality (even a mistaken one!) getting special news coverage... [https://globalnews.ca/news/7326819/alberta-coronavirus-covid-19-september-10/]

In another upsetting observation, these ages are expected to be most affected - in a positive way - by "excess survivability" effect. Pandemic has removed many "traditional" causes of (mostly accidental) deaths among younger people - so there should have been fewer young deaths, not more!

Incidentally, a very similar research of the excess mortality in the USA (https://www.cdc.gov/mmwr/volumes/69/wr/mm6942e2.htm) also shows new disproportional mortality in younger people - so this is not something Canadian-specific. It seems that younger people are indeed not immune from the deadly dangers of the pandemic - but not in a way one would think of...

Elderly Mortality

Sharp reduction of elderly mortailty is another puzzling effect after youth mortality rize. With adjusted baselines, this drop is expected - but not to such extent. Even all-Canada excess elderly mortailty is not impressive, and in many jurisdictions it hovers at mere percents, occasionally going below zero. Yes - in some places excess mortailty is negative in the older age groups!

One possible explanation comes to mind. When any COVID deaths are announced, we subconsciously imply that all these people would have still been alive, if not for the virus. But this is very much not the case! Vast majority of direct virus fatalities occur in the same age groups that are already dying out - actvely and naturally, mostly from simply "being old". Low or negative excess numbers show that in such places majority of COVID-caused deaths are not "premature": if not for COVID, those same people would have likely die within 2020 anyway.

Regional Diversity

Another major finding is the significant regional diversity across Canada. Divides in demography, economy and lifestyles between provinces and regions are so deep, that they should be compared with much care. The graphs above show, that, of all provinces, only Ontario resembles the cumulative Canadian data close enough to be useful. If you live anywhere else - forget any all-Canadian stats and use your local numbers and situations. Not only numbers vary across Canada - the very date ranges to which data is applied can be different!

Canada's slow and archaic data collecting system is further impended by provinces. You'll often see graphs like this one - weekly plot of all excess deaths extracted from the all-Canada table:

Data

However, a deeper look reveals that the data, on which this plot bases, is not uniform across provinces, and misses some important elements. This is how data looks mid-February 2021:

  • Alberta, British Columbia and Newfoundland data is good up to and including November;
  • Manitoba, Saskatchewan and Nova Scotia data is good up to and including October;
  • Ontario data (which is the biggest chunk of data here) is good up to and including September;
  • Quebec data is generally good for the full year - except for the most intriguing 0-44 age group which completely stops after week 35 in August (!)
  • Yukon data is missing for full three past years, impossible to calculate baseline;

We don't really look at smaller provinces (except for Yukon which "jumped out" by completely failing data verification). Remember of these limitations next time you see a diagram like this! They do not invalidate data, but remind us that Canadian statistics is not only painfully late, but also non-uniformly late.

Back to Top ↑

Graph Explanation

Now let's look at different provinces, starting East to West, and skipping really small "bubble" provinces. Here is a breakdown of what the values are in the diagrams below:

For smoother look, each data point combines together four weeks.

Back to Top ↑

Provinces - East

Quebec

As explained earlier, we have to truncate Quebec data after August due to missing numbers on 0-44 age group - making it the shortest date range among provinces:

QC

Quebec is unique in many ways. More than one feature in its diagram is not seen anywhere else:

Ontario

Ontario data go for a whole month longer than Quebec:

ON

In the first check, Ontario data was repeating All-Canadian data in such close resemblance, that it is tempting to plot the whole Canada next to it and compare. Of course, all-Canadian plot must be limited to August, otherwise incomplete Quebec data will invalidate its trailing section:

CAN

Indeed, close enough! If you forget absolute numbers and look at more "generic" percentages and ratios - the plots and tables are almost twins!

What is not pleasant about this graph - after a rather ideal Quebec - is a gradual increase of mortality percentage from older to younger groups. Also, youth mortality increase, while being highest among other groups, looks quite modest with only 6:1 non-COVID to COVID ratio.

Canadian ratios look a bit worse, of course, likely due to Western influence.

Back to Top ↑

Provinces - West

First, let's look at all four western provinces joined together:

West

"Combined West" looks very different from both the "whole Canada" and any Eastern province in particular:

Manitoba

Good Manitoba data goes to the end of October:

MB

Absolute numbers of excess mortality in Manitoba are not that big to make statistically significant predictions. But even with these numbers you can see that almost all excess fatalities were younger people dying "from other reasons". That does not look good at all!

Saskatchewan

Good Saskatchewan data also goes to the end of October:

SK

Like in Manitoba, Saskatchewan's excess deaths are almost exclusively collateral. Unlike in Manitoba (and unlike anywhere else) their weight is unproportionally shifted from younger people to middle-aged group. Those are absolute numbers - percentage-wise and rate-wise, younger group is, unfortunately, the absolute leader, like everywhere.

Alberta

Good Alberta data goes to the end of November:

AB

Uniquely for Alberta, the youngest strata had suffered most in every single aspect - absolute numbers, percentages and ratios altogether. More than in any other province! No wonder so many young people are so unhappy here!

Excess mortality in the oldest strata is bigger than in the east - but not by much. Same as everywhere else: people killed by COVID in these ages would have been dead anyway - from natural causes.

British Columbia

Good British Columbia data goes to the end of November:

BC

BC picture is generally similar to Alberta. Only so slightly less horrible for youth strata, just a little worse for older ages. Between them, it is easy to miss a bigger non-COVID hit to the middle-age strata... but it is there. In general, COVID mortality is fully covered by expected baseline, so non-COVID share is prominent in every age group.

Back to Top ↑

Atlantic Provinces

As a bonus, let's have a quick look at two Atlantic provinces - Newfoundland and Nova Scotia. They are both too small for any meaningful detailed stats, but maybe we can spot some differences ant least in the trends:

NS
NF
Nova Scotia Newfoundland
NS NF

 

Back to Top ↑

Final Thoughts

It is hard to add anything to the numbers and graphs - they speak for themselves, and their story is a sad one. Just a few findings are worth reiterating.

The best finding is that the virus is much less deadly than many scared people still presume. At the same time, it is a little more deadly than "other many people" think. This study does not address the question "are COVID-fighting measures killing more people than they are saving?" - because we don't know how many we are saving.

This study answers a different question: "are COVID-fighting measures currently killing more people than COVID itself?" And the answer depends on where you live and what ages you consider.

For numbers for all ages in big Eastern provinces do not show any non-COVID collateral damage. "Combined West" reveals collateral deaths at 3:1 ratio.

In different age groups, things are much more diverse. Younger people face unusually high collateral mortality increase even in the East. This effect is hidden when data is reported for all age groups. In the West, the growth in the youth mortality is so staggering it shows even in the overall numbers!

Older people, on the other hand, often do not see the excess mortality at all - COVID deaths simply replacing natural causes. And where there is excess elderly mortality, it is seldom accompanied by collateral side effects - West being a sad exception once again...

This study highlights the drawbacks and really unacceptable time lags and quality gaps in Canadian statistics collection, especially compared to other countries!

This study verified the expected divide between East and West. While East had been harder hit by the actual virus, West suffered a lot more from the collateral chaos.

What can be done about all this? Not much. It will take something more drastic than a couple of long-suspected revelations to change anything in the pandemic response. Governments already played every card in their sleeve, and all we can expect are the various combinations of repeating the same - each time with more desperation.

But with all this available data coming, we should at least study, study and more study - every bit of it. If you can see the problem that others can't - you may be able to at least better plot your own personal path through the muddy waters.

  1. There are many good political reasons to concentrate on mortality numbers:

    It is the main divisive point between the two social camps.

    Most heated arguments between the hard-liners on both sides are burning on the "this is just another flu" - "no this is not" fault line. Most urging motivators stick to "saving the lives" mantra, most stubborn resistance comes with "this is not nearly as deadly as they say" notion. We may be able to answer this important question by quantifying the arguments.

    It is the most visible political agenda

    Modern society sees human life as an ultimate value that trumps any other values. This new "safety religion" indeed makes our day-to-day lives generally safer - but, like everything, brings its own unique "dark sides" and "cracks to fall through". Safety religion directly affects every aspect of COVID pandemic response. For example, no government can afford to accumulate too many too visible fatalities too quickly. This becomes totally "politically unacceptable" even for the harshest political regimes, not just for the first-world democracies. Visible ability to "save lives" becomes major political asset, visible failure to do so is a sure political disaster.

    It is the most emotional topic personally

    Emotion toll of any death, expected or not, is high. When looking at the this data, remember that every data point in this table is a tragedy, someone's passing away, leaving grieving family - which, due to pandemic, is denied ability to hold a hand of a dying relative, to gather together for a last farewell, a decent church service, celebration of life and proper funeral. For many of us this experience is recent, close and personal - sincere condolences to anyone who had to go through this in 2020...

  2. Mortality stats is also most beneficial for its data quality and accuracy:

    It is the most complete

    Unreported deaths are extremely rare. Death is a serious, irreversible, tragic accident, and is thoroughly documented and sometimes deeply investigated.

    Other stats - like daily "case numbers" - depend a lot on factors like tests availability, eligibility, tester qualification, lab capacity, turnaround times, test reliability... They are also the easiest to be affected by people's behaviour patterns - when people try to avoid testing for various reasons. So daily number of "465 new cases" can easily have a +300-30% error on it.

    It represents a short event

    Terminal outcome is always assigned a definitive single time stamp - even if the exact date is uncertain.

    Hospitalizations and ICU occupancy (better criteria than case numbers), still tell more about the health care system capacity and quality than about epidemiology situation. They are multi-day (or multi-week events), harder to be attributed to a particular date. Their counts may be affected by non-COVID patients requiring same beds or ICUs, or by people choosing to stay at home even with severe symptoms. Also, the tone in the news for these is very elastic on supply: jurisdiction with more resources will be more relaxed reporting the same numbers than a similar but less-prepared jurisdiction.

    It is immutable

    There may be occasional early corrections for date and cause of death. But, as a rule, daily count of 27 deaths will go down in history for that day as 27 and will remain 27 for as long as anyone cares (and then for some time after).

    It is the most tamper-proof

    Death event is a serious business anywhere on the planet. It leaves numerous paper and digital traces, making it next to impossible to intentionally manipulate records even in the most obscure political regimes. This makes it an ideal comparison criteria for regions that are otherwise totally incomparable.

    Of course, first-world countries don't falsify statistical data (and third-world countries might not even have any data to falsify). But our planet is mostly occupied by all kinds of shabby second-world regimes doing whatever they wish within their boundaries.

    Data manipulations go both ways. Some regions hide COVID mortality to report success on fighting the pandemic. Others over-inflate the cases to demand more funds to fight them. Foreign researchers often see 1000% discrepancies in regional data - so they resort to the "ultimate truth" of death records to get down to the grim reality.

 

author
Back to Top ↑

 

Facebook Reddit LinkedIn Twitter