On numerators and denominators

Have you had the Coronavirus yet?

How do you know? Have you been tested and, if so, do you believe the test results?

The reason I ask is because I have a theory.

It’s not a passionately-held theory, so don’t read this as a call to reverse public policies and send everyone back to work, school and mass public entertainment events.

But it’s a theory for which I am increasingly open to exploring further. It is this:

COVID19 is far more widespread and, on a per capita basis, far less deadly than the current public estimates.

What would we need to know to be able to disprove or confirm this theory?

The following data points for a suitably large population:

1. The population size,
2. The number of people who have had or who currently have it, and
3. The number of deaths attributable to COVID19.

Simple, right?

Not so fast. (1) is easy enough, we have a good handle on how many people live where,

Collecting (2), though, has three significant problems.

Firstly, it currently relies on the results of a test that has not been universal or even widespread in its use – many people displaying symptoms have been turned away when requesting tests for not “fitting the criteria”.

Second, the test is “real time”. i.e. it tests whether you currently have the virus, not whether you have had it and fully recovered.

Thirdly, the ratio of false negatives is not known. How many people have been sent away with “just the ‘flu” when they’ve actually had the Wuhan version? We don’t know.

Finally, we can’t be certain of the death rate from COVID19. Different countries are counting it in unique ways. The death of someone with stage four cancer in one country may or may not be attributable to COVID19 depending on the methodology.

These complaints about data may sound like nitpicking but they make a significant difference to how we might respond at a national level. Or at least they should.

At the risk of stating the blindingly obvious, the problem is the lack of a quick and accurate antibody test, one that would tell us who has been infected and recovered (and presumably, has immunity).

There’s good news though, it would seem one has been developed in the USA and is hopefully on its way to your local population.

I wonder what it will tell us?

Bill’s Opinion

I’m very sceptical of multi-variable predictive models. Minor inaccuracies on the estimate of one or more variables result in wild ranges of outcomes.

I’m not denying COVID19 is a massive public health problem though; even if it isn’t as deadly as the models are suggesting, the contagion rate is rapid enough to swamp our health systems when it hits people with vulnerable immune systems concurrently.

I do wonder though whether many of us have already had it in early January and shrugged it off as a bad cold. If that describes you and you’ve since lost your job, I’m doubly sorry for you.

Once this test has been distributed widely to a sample population, perhaps we’ll know whether the models were accurate or not.

Lastly, I’m in good company.

UPDATE; This is an excellent analysis of the statistics.