Understanding the Numbers Behind COVID-19

Predicting an epidemic requires understanding a few simple metrics.

As you know, in this forum I usually talk about a new study with a focus on the data and its broader implications. This week, I want to take a bit of a deep dive into how infectious disease epidemiology works, using, of course, the novel 2019 coronavirus as a great real-time example.

You’ve heard a ton about the new virus over the past month – reports in top medical journals detailing case series, breathless newscasters asking if this is the next Spanish flu, and of course some cautious statements from government officials charged with containing the pandemic.

All of those reports focus on numbers – cases, incubation periods, attack rates, fatality rates, basic reproductive number. But I think as healthcare providers we need to have some better intuition about what these numbers really mean, how they fit in with other infectious diseases that are more familiar to us, and importantly – how they can be misestimated. Because the vagaries in these estimates can make the difference between a flash-in-the-pan scare and full-blown worldwide panic.

Ok let’s start with the big one – the basic reproduction number, or R0. This represents the average number of susceptible individuals an infected person will transmit the disease to – think of it as contagiousness.

How R0 works

There’s two major drivers of R0 – the number of contacts infected people have while they are infectious, and the attack rate – or the percent chance a given contact will get the disease.

Logically, if the R0 is less than 1, a disease outbreak should wane over time, and if it’s greater than 1 cases should continue to increase. Seasonal flu, for example, has an R0 of around 1.5. The Spanish influenza of 1918-1919 had an R0 as high as 2.  Chickenpox, which is fairly infectious has an R0 of around 5.

Right away we need to notice something. The R0 is clearly not the measure of how terrible a new infection will be. The Spanish Flu killed 50 Million people in 1918. I’ll take chicken pox over Spanish flu any day of the week.

Enter case fatality rate – the percentage of infected individuals who die from the disease. Historical reports put the case fatality rate of Spanish flu as high as 10%. SARS had a case-fatality rate in that range, incidentally.

So if you want to predict how terrible a new disease will be, you really want to look both at the basic reproduction number and the case-fatality rate. Fortunately, evolutionary processes tend not to favor highly fatal diseases – after all, a dead host doesn’t transmit more disease to others.  The exceptions include some of our most feared conditions.

HIV before treatment was available – with an R0 of around 6 globally and a near 100% mortality rate. Smallpox, R0 of 5, mortality rate of 30% in the unvaccinated. Bubonic plague. R0 of 3, untreated mortality rate of 60%.

The worst of the worst live in the upper right

Where is the new coronavirus in all of this?

Well with an R0 of 2.5 and a reported case-fatality rate of around 2%, this could be a major problem.

But that 2% is probably wrong.

Remember, the case fatality rate is defined as the number of fatal cases divided by the number of total cases. We probably capture fatal cases accurately – people who are that sick generally end up in hospitals. But we may be missing the number of total cases by huge margins – perhaps even an order of magnitude, because asymptomatic and mildly symptomatic people may not be getting tested. If this is the case, we should see the case-fatality rate decrease as screening improves.

We can also change the R0 by addressing those two elements inside it – the number of contacts an infected person has, and the attack rate of the disease.

Limiting potential contacts can be achieved through isolation and quarantine.

Attack rate can be reduced by wearing masks, handwashing, and of course, vaccination were a vaccine to become available.

The wrinkle here is that these interventions depend on identifying cases, and it is still an open question as to whether transmission can occur in the asymptomatic period.

According to Dr. Anthony Fauci, Chinese officials have stated they have evidence of asymptomatic transmission – which is bad if true – but how bad depends on another epidemiologic factor – the latent period: the time between infection and first symptoms. For the new coronavirus this appears to be around 5 days – not too long.

That’s good. A combination of a long latent period and infection that can be transmitted when symptoms aren’t present is a recipe for disaster – see HIV again for a dramatic example of this.

Ok there’s a lot about Coronavirus that we don’t know – but I have to say I have been impressed by the speed of the scientific community in narrowing in on not just these key numbers, but even on the key other elements of infection control – identification of possible treatments, and development of vaccines. The eyes of the world are watching the numbers, and hopefully now the numbers are a bit more meaningful.

Wash your hands.

This commentary first appeared on medscape.com