Handling uncertainty: panic vs. precautions…

It’s always hard to make the right decisions when the data is uncertain. And that’s currently more important than ever, faced with the rise of the SARS-CoV-2 novel coronavirus and COVID-19, the disease it causes.

For example, there are currently many discussions about the “true” mortality rate of the Covid-19 coronavirus.

Some people insist that the currently reported mortality rates are wildly overestimated because of the “missing denominator” problem—the absence of people who don’t show symptoms. If this number were sufficiently high, the novel coronavirus could be “no more deadly than the flu” (although, of course, the flu is a big killer: over half a million deaths annually around the globe).

Others point out that seems to be an optimistic assumption based on the currently available numbers, and that a higher hospitalization rate (even if everybody ultimately recovered) could alone lead to serious strains on healthcare resources (and hence negatively impact the survival rates of other types of health problem).

Without the right numbers, how do we make sensible decisions about how serious the threat is and what precautions (or as one analyst called it “self-inflicted damage to the economy”) to put in place ?

Unfortunately, like so much of analytics, it turns out that there are lots of unknown values and nuances involved in calculating a definitive number. For example, there’s typically a several-week delay between somebody getting the disease and suffering severe symptoms, so if cases are growing over time, a naive mortality rate (cases divided by deaths) would underestimate the real value (since some of the cases currently counted may eventually be mortal).

Researchers, of course, try to use sophisticated statistical techniques to get around these problems, and have attempted to provide their best estimates for outbreaks around the world. But the numbers appear to be very different across regions—because of wide variations in age, susceptibility, and treatment methodologies of different populations.

A more flexible way of attacking uncertainty is to look beyond specific models and instead benchmark against “other people like us.” Here, for example, are the curves of discovered cases across several countries, all showing remarkably similar trends.

If the growth of new cases is exponential, the danger is that precautions will seem alarmist until they suddenly seem insufficient—and that having low numbers today is not, per se, a good reason not to to act. If a fire breaks out in the kitchen, pointing out that the rest of the house is just fine is not a convincing argument.

I think all of this is a good example of where talking about detailed numbers or comparisons can obscure rather than reveal what’s important (e.g. “only 0.0004% of the population has this!”, “I’m more in danger driving in the rain!”).

Governments around the world are instead asking themselves “does the progression in other countries seem to indicate that, if we don’t take precautions, there is a danger of overloading health systems, and endangering the lives of the most vulnerable?”—and are concluding the answer is “yes”.

Given that countries like Italy (and China, Singapore, Korea, etc.) have been forced to implement more draconian quarantine restrictions over time (closing schools, travel bans, etc.), it seems reasonable that the real question for other regions is “do we have any reason to believe that the trend will be any different for us?” If not, those countries should probably take common-sense “social distancing” precautions earlier than they might otherwise have planned to do so.

It is clear that for most of us, the odds of having serious health problems because of the coronavirus are low, and so there’s absolutely no reason to panic. As much as possible, we should target measures to avoid pain for other vulnerable groups, including minimum-wage workers and others who don’t have the luxury of just sitting this out.

But the flip side of scare-mongering is “complacency-mongering”. Low personal danger shouldn’t be confused with a low risk for the world as a whole, and we should all take our role in slowing down the epidemic very seriously (in order to smooth out access to health resources and increase survival rates, see below).