Thread by @MoriartyLab

I think that it might be really helpful for journalists, and Canadians in general, to know how to find out for themselves how many deaths from any cause have been reported in Canada since the beginning of COVID-19.

I'll use Saskatchewan as an example for how to do this.

First, let's start with a view of reported COVID-19 deaths to date in SK, provided by the incredible volunteer team of SKers Noah Little and @covid_canada. They have been collating Canadian COVID statistics for the country since Mar/20.

These are new daily SK deaths over time.

@covid_canada Next, you need to know where to find deaths reported to the Canadian Vital Statistics Death Database.

It's available courtesy of @StatCan_eng which does a LOT of working ensuring that much of the data they collect is publicly available.

Here's the link: www150.statcan.gc.ca/t1/tbl1/en/cv.…

@covid_canada @StatCan_eng The table is called "provisional" because death reporting is not yet complete. It's important to remember that these numbers will keep going up until death reporting is complete.

In Canada, it takes 2 or more years for death reporting to be completed.

@covid_canada @StatCan_eng For example, Saskatchewan death reporting for the week of Feb 1, 2020, is still only 95% complete. As of Oct 1/2021 SK death reporting is about 70% complete. No deaths have been reported at all by SK after Oct 1.

@covid_canada @StatCan_eng So keep in mind that the numbers you see here are LOWER than actual numbers of deaths, and will be until 2 or more years after the date you're interested in looking at.

@covid_canada @StatCan_eng OK, to know where death numbers see "normal" or not, even in a ballpark kind of way, you need to know how many deaths happened per year in previous years.

For SK, on average 9,300 people died per year from 2015-2019.

Knowing this, let's start by adding up the deaths in 2020 (remembering that reporting still isn't done yet and that 2020 deaths are still being added).

How many people died in Saskatchewan in 2020?

It was 10,580.

What is 10,580 minus 9,300?

1,280. So, there were about 1,300 extra deaths in 2020 compared to the average for previous years.

Now, let's do 2021.

8,415 deaths have been reported so far for SK in 2021. That's because reporting is not yet complete.

Not only will new deaths still be added for most weeks in 2021 for another 1-2 years. There also haven't been any deaths reported at all from October 1 on.

OK, so SK has reported 8,415 deaths for 47 weeks of 2021. That's about 180 deaths/week.

From Oct 1 to the end of Dec/21, SK was having a pretty serious COVID death wave. It was about as big as the 2nd wave (big).

Shoot. I made a mistake. Jan 1-Oct 1/21 is not 47 weeks. It's 39 weeks. So, about 220 deaths/week up to Oct 1. Multiply that by 13 weeks, and you get 2,860 deaths. Add that to 8,415 and you get minimum 11,275 deaths for 2021.

Subtract 9,300 (avg 2015-2019) and you get 1,975.

1,975 plus 1,280 equals 3,255.

So, 3,255 extra deaths in SK during the COVID-19 epidemic to the end of 2021. Keeping in mind that this is a big UNDERESTIMATE because of incomplete death reporting.

Because SK has not reported any deaths at all after Oct 1/21, this means that it's only completed reporting for 60% or less of all deaths during the COVID epidemic.

You can see why slow, incomplete death reporting is a problem, especially in the slowest provinces, BC and SK.

The other thing you should look for is lower than expected death numbers compared to 2015-2019 in any given week. This is more work and you need to download the spreadsheet of data to do that.

If you do this you can see that reported deaths are lower than expected in multiple weeks, especially considering that there were quite a few COVID deaths being reported in these weeks.

Estimating actual excess mortality is more complicated than this. You have to figure out if increases are statistically significant, and when I do excess mortality estimates I also correct for toxic drug deaths.

You also need to take into account that deaths will increase a little bit each year due to population aging (although that doesn't affect SK as much as QC and Atlantic provinces).

Finally, you need death reporting to be at least 95% complete in the province you're interested in.

In fact, by watching Quebec data really closely through the epidemic, I've observed that death reporting needs to be >=99% complete to get really good estimates of excess mortality.

But despite all these caveats, if you want an estimate of the MINIMUM untimely (extra/excess) deaths in SK to the end of 2021, you can make ballpark estimates like this.

They'll go up over time--a lot in places like SK that haven't yet finished reporting for the first wave.

How many COVID-19 deaths were reported in Saskatchewan to the end of 2021.

955.

I'm happy to do a Zoom tutorial to show people how to do this, if that would be easier than reading through on Twitter.

It's important for all of us to be able to find, understand and interpret our provincial and national statistics.

Yes, it's a fair bit more complicated than I've described--but if you want to look things up to gut-check things you're hearing, this is probably good enough.

Just don't forget! These are UNDER-estimates. They'll increase considerably over the next 2 years.

Finally, that 7-fold estimate of under-reporting that some have dismissed as impossible. It comes from our Royal Society of Canada report, and comes from a period in 2020 when Saskatchewan death reporting was considerably more complete. It was <95% complete, but we had to use it.

That under-reporting estimate was done with excess mortality estimates adjusted for toxic drug deaths, and used only excess deaths in people 45 and older, to be as conservative as possible.

By July/22 I think we'll have the final SK estimate for Wave 1.

And so it goes.

P.S. I'm really tired right now. The estimate I provided here is lower than one I did earlier today, and I think I've missed a few weeks somewhere, but don't have energy to go back and check.

Maybe you can try it out and do some error checking for me! It's a good way to practice