When is a death not a death?

On the 12th of August 2020, Public Health England announced a new definition of a Covid-19 death. What’s the issue?

The original definition was:

“A death in a person with a laboratory-confirmed positive COVID-19 test and either died within 60 days of the first specimen date or died more than 60 days after the first specimen date, only if COVID- 19 is mentioned on the death certificate”

And the new definition:

“A death in a person with a laboratory-confirmed positive COVID-19 test and died within (equal to or less than) 28 days of the first positive specimen date”

Adopting the new definition led to an overnight drop in the UK Covid-19 death toll of more than 5,000. And it also led many to questions over what exactly is being counted.

Covid Dashboard highlights the latest stats from public authorities detailing the health and economic impacts of Covid-19

Do not use

Both definitions are non-specific and leave room for doubt – someone who who had tested positive for Covid-19 but died from another cause is counted, while someone who may have died of Covid-19 but not been tested is not counted. 

It’s easy to get bogged down in an intellectual morass here. I think the most fruitful point of view is pragmatic: what exactly are we trying to achieve by tallying deaths? I would wager the main thing that people are interested in is answering the question of how many deaths are “caused” by Covid-19. Even then there’s still a long way to go – what exactly does it mean for death to be caused by Covid-19?

There are many smart people who have thought about such things a lot, and roughly speaking they have converged to the following answer:

Imagine we ran two parallel versions of the universe. These universes are identical, except in one there is the novel SARS-Cov-2 virus, and in the other there is not. Think of the inhabitants of Earth in the second scenario as being a type of placebo control group; we set up the experiment such that they are equally convinced of the reality of Covid-19, and they respond just as we have – with a raft of restrictions, lockdowns, and other policies. It just so happens that in their universe, there ain’t no such thing as Covid-19.

Is this actually possible? Of course not. But it’s our thought experiment, and in the Platonic realm of pure thought we can do whatever we want. The point is that if we could somehow figure out how many people die in the Covid-19 placebo universe, then we could simply subtract that from the number of people who die in our current universe, and it would be eminently sensible to attribute the disparity to deaths caused by Covid-19. We did a controlled experiment where only one thing was changed, so any disparity can be attributed to that one thing.

So, there are people who actually do try to figure out how many people die in the placebo universe (they call it a “counterfactual”). You can imagine how difficult that is. In practice it involves some very esoteric mathematical modeling and statistics. You can find some information about how the CDC do it here.

Of course, such efforts are inevitably flawed – the task is an impossibly difficult one. However, for now, such efforts appear to be the most sophisticated approach we have to counting deaths caused by Covid-19.

And the moral of the story is to be very skeptical of interpreting widely reported death tolls as indicating the number of deaths caused by Covid-19.