Excess mortality and COVID-19

Some have worried that COVID-19 cases are being over diagnosed. This almost certainly isn’t true. A useful tool is the excess mortality tracker that measures how many more deaths there ate currently compared to past years.

I’ve posted on this before, but there’s now a running total tracker here.

Continue reading

Real world data on COVID-19 antibody tests – we ain’t there yet

We’ve talked here about the problems and limitations of antibody tests. We’ve seen others discuss it. And a primer on the mathematical whys and wherefores is available here.

Today some pre-publication data on a variety of US antibody tests were released, and they aren’t great–and they illustrate precisely the problems discussed above:

Continue reading

Human and economic costs of COVID-19

As we continue in lockdown, the costs mount in human and economic terms.

One problem which we’ve noted frequently is that it is difficult to get a true sense of the virus’ fatality rate, since knowing how many cases there are depends very much on widespread testing. Do less testing, and you’ll find less virus. But if you find less virus, then the fatality rate (% of people with the disease who die) will appear higher than it truly is. (And, since we’re unlikely to ever measure every single case, fatality rates should be treated as a sort of “upper bound”–the actual rate will almost always be at least somewhat lower.)

I ran onto an interesting and useful way to think about such things, published in The Economist this week.

It takes the approach of calculating “excess deaths.” That is, with historical statistics you can know what the typical rate of death at any time of the year is from all causes. Those numbers might fluctuate a bit (say in a bad flu season) but from year to year in a given country, they stay remarkably consistent.

So, one approach to getting a handle on COVID-19 is to simply say, “What’s the excess rate of death?” How many more deaths are we having this year (regardless of diagnosis) than is typical?

The data is mostly from Europe. Here’s their graph for the aggregate statistics compared to the last 10 years:

You can see the peaks of the last few years’ influenza (2017, 2018, 2019). Those have been exceptionally high years compared to the average death rate over the last ten years:

Compared to the baseline average of deaths from 2009-19, the flu seasons of 2017, 2018 and 2019 were all unusually lethal. But the covid-19 pandemic, which arrived much later in the year, has already reached a higher peak—and would have been far more damaging without social-distancing measures. Compared with the baseline, EuroMOMO’s figures suggest that there were about 70,000 excess deaths between March 16th and April 12th.

This also allows them to compare the known COVID-19 deaths with these total excess deaths. So, you can get a sense of how good the testing is. In this graph, New York does very well (but one wonders if there are undiscovered deaths at home, etc., that may change this) compared to some third world countries who understandably do much worse (though the first world does not necessarily cover itself in glory either). One wonders too if New York is including deaths as a result of COVID from other causes in their stats:

Why is this useful?

I think that this has a few benefits as we try to understand COVID-19:

  1. it is less likely that officials would miss a death than miss a COVID-19 diagnosis. Deaths are easily diagnosed with the naked eye!
  2. It lets us access some of the “hidden deaths” that might not be COVID infections, but result if a healthcare system is overwhelmed. [These are, say, your heart attack patients who would have lived if there’d been an ICU bed, or if they weren’t afraid to go to the hospital, or if the ambulances weren’t taking 2 hours to get somewhere instead of 5 minutes.) An effort to lower these “hidden deaths” (which could affect any age group) is one big reason for the “social distancing.” So, this type of analysis can be used to determine what the costs of a health care system overload actually are.]
  3. It could also get you a look at causes of death from lockdowns if you picked the right time window. One potential issue could be higher rates of suicide, for example, compounding because of either lack of social supports, financial strains, or fear. This is a complex subject, which I’ve treated in a separate post here.
  4. It gets us a look at impacts on a variety of age groups. Here’s the same data separated out by age:

So, while the over-65 group (the top graph) certainly has an elevated death rate, people younger than 65 are also markedly increased. So this tells us that COVID-19 is not simply hitting old people in extended care homes–there is a cost to younger people too.

Interestingly, the 14 and under group has a lower death rate. This makes sense–COVID-19 is rarely fatal to children, and lockdowns mean that kids are doing fewer “risky” things, like riding bikes, or getting hit by cars, or driving around in cars.

This also highlights that we always accept some degree of risk. Our kids would be safer if we locked them in the house forever. But, we clearly don’t and shouldn’t do that. Some risks are worth running.

Defining where that line is for COVID-19 is one of the vexing problems currently facing citizens and governments.

I’d like to see more analysis like the above, particularly focused on Alberta and Canada–it might guide policy-makers in these types of decisions.

Estimating economic value of a life

An interesting, if somewhat cold-hearted, look at economics vs epidemiology: Globe and Mail.

In this example, they don’t mention adjusting for age of death, which is obviously going to be a factor in the economic “cost” of each death (aside from the obvious human costs, which I trust don’t need to be spelled out).

Bayes’ Theorem for people who don’t remember much high school math

Yesterday, I posted a link to a great discussion about COVID-19 testing, and the various types of tests that do or will exist, and their strengths/weaknesses.

It can strike many people as strange that doctors’ tests aren’t 100% accurate. (It even surprises some doctors!) No test is perfect.

But, not only are tests not perfect, but even a test that sounds like its very accurate can be misleading if used under the wrong conditions.

Continue reading