Thread by @MoriartyLab

Concerns about possible under-reporting of #COVID19 deaths in 4 #Canadian provinces:

Ontario, Saskatchewan, Alberta, BC

This thread will be fairly long. I've assembled it to try to crowd-source insights from those knowledgeable about provincial C19 death reporting.

First, I'll start with plots of cumulative excess deaths in Canada until Nov 14 (the last date when we have fairly complete excess death reporting for most of Canada).

I've set the start date as the first week when a province or territory reached 1 reported C19 death/million pop

Here are the excess deaths/100K, for those interested.

Most of Canada's excess deaths during #COVID19 epidemic have been in people older than 44.

Excess deaths in 0-44 group might be C19-related, but are more likely due to parallel opioid death epidemic, since most opioid deaths happen in people 20-49.

Suicide has NOT increased.

Following are a series of graphs of excess deaths broken down by age and sex for different regions and provinces.

They're very interesting in their own right, and I know people will want to see them.

Atlantic provinces and Northern territories: Excess deaths, by age and sex

Pooled data are presented for Atlantic provinces and Northern territories, where #COVID19 epidemic has been well-controlled

This assists comparison to other regions with more poorly controlled epidemics

Quebec: Excess deaths, by age and sex

Ontario: Excess deaths, by age and sex

Manitoba: Excess deaths, by age and sex

Saskatchewan: Excess deaths, by age and sex

Alberta: Excess deaths, by age and sex

British Columbia: Excess deaths, by age and sex

Here are total reported #COVID19 deaths over the same period.

There are two things to note:

1. There were about 3,000 more excess deaths than reported C19 deaths by Nov 15.

2. The provinces look quite different in excess death and reported C19 death graphs (see next tweet).

Here are total excess deaths in people 45 years and older.

ON, BC & AB look quite a bit higher than in the reported C19 death graph in the tweet above, and QC looks considerably lower.

In the following tweets, I'll compare reported #COVID19 deaths to excess deaths in people 45 years and older to try to figure out some of this.

First, Atlantic and Northern Canada

Excess deaths were less than reported #COVID19 deaths.

Excess deaths are deaths on top of what we expect normally. Some extremely frail people who got C19 might have died during this period anyway, so excess is lower than reported C19 death.

Now Quebec, where we see a similar pattern, but writ very large. Reported #COVID19 deaths are greater than excess deaths--about 2000 deaths greater.

It's likely this is partly because of how many C19 deaths occurred in QC nursing homes, more than any other province or region.

In Manitoba, excess deaths in people 45 years and older was close to reported #COVID19 deaths, and most excess death was in younger people.

But note: the excess death data end just as C19 cases and deaths in MB were really taking off. It will be interesting to see later data.

Now is where things get interesting: ON, SK, AB, BC.

Before we get to this, some important things to know.

1. The WHO recommends that all countries report both probable and confirmed #COVID19 deaths.

2. Canada's probable C19 case definition for surveillance is tricky.

Cont..

Definition of a probable case in someone who has not been tested:

1. Person has symptoms compatible with #COVID19

AND

2. Had high-risk exposure w/ confirmed case OR was exposed to known cluster or outbreak

You can see where this is going....

So,if someone with C19 symptoms dies but is not tested and is not epi-linked to outbreak, cluster or other case, then this is not reported as a C19 death.

In congregate care outbreaks, C19-like deaths in untested people can be reported, but in community they may not be.

It's also likely that deaths in people in community by stroke or other less stereotypical C19 symptoms might not be recognized or reported as C19 if the person didn't know they'd been C19 exposed, and this would be less likely in settings where contact tracing has collapsed.

There could be many other issues with reporting at play too. For example, there have been media reports that some provinces, including ON, don't report probable cases (and deaths?) as part of their daily reported numbers. I'd like to know if this is still true and if so, where.

And of course, the more undetected community transmission you have the more likely that you'll have unrecognized or untested C19 deaths (although QC is the contradiction to this hypothesis).

OK. Back to the remaining provinces.

Ontario

Unlike in Atlantic and Northern Canada, QC and MB, excess deaths in people 45 years and older are 50% higher than reported #COVID19 deaths.

Why? Has ON missed 50% of C19 deaths, or does the province report differently than the regions above?

Saskatchewan

Excess deaths in people 45 years and older is 10X higher than reported #COVID19 deaths.

Why?

It's not opioid-related deaths (different trend than in people 0-44 years).

It looks like SK missed A LOT of C19 deaths up to Nov 15.

Alberta

You can really see the likely effects of the opioid crisis on excess deaths in younger people.

But excess deaths in people 45 and older are still 2X higher than reported C19 deaths.

The numbers were close till early September. Did AB start missing deaths after that?

Finally, British Columbia.

This one is really heart-breaking. It's a big province with more public health resources and expertise than SK.

But excess deaths in people 45 and older are 8X higher than reported #COVID19 deaths. And they've been higher since the epidemic start.

In closing, here are side-by-side graphical comparisons of provinces showing:

1. Excess deaths in people 45 and older/100K people
2. Reported #COVID19 deaths/100K people

They tell two quite different stories of the impact of the epidemic up to November 15.

Comments welcome.

Here's a viewable link to the spreadsheet for those who'd like to look at the raw data and calculations:

docs.google.com/spreadsheets/d…

/end