Contagion Number, Explained
The
coronavirus, the epidemic(?) that’s been spreading in (and from) China. But how
contagious is it exactly? Ed Yong wrote this superb article explaining the number
used to measure such things in general. It is called the “basic reproduction
number”, or just R0, pronounced R-nought:
“R0 is the average number of people who
will catch the disease from a single infected person, in a population that’s
never seen the disease before. If R0 is 3, then on average every case
will create three new cases.”
In case
of the coronavirus, 6 research teams have come up with a value for R0, spread
over a wide range from 1.4 to 5.5. Why such a discrepancy in values? Because in
the early days, there’s just not enough data:
“Some people might have been infected
without showing symptoms. Others might not have reported their symptoms to
health authorities.”
Unfortunately,
a Harvard doctor seized upon the R0 value of 3.8 from one of those studies and
called it “thermonuclear pandemic level bad” in a tweet. Yong counters that
interpretation by pointing out the R0 values for SARS (2 to 5), HIV (also 2 to
5), and good old measles (12 to 16).
Secondly,
Yong says a high R0 number doesn’t mean it will spread fast:
“It is a measure of a disease’s potential. And once nations realize that
a new disease exists, they can actively screen for it, check that health-care
workers are using proper protection, and instigate quarantines. Even simple
steps such as hand-washing might make a difference. All these measures could
potentially lower the chances that the virus will spread and ensure that
its actual transmission rate—the
quantity known simply as R—is less than R0, and ideally less than 1.”
Flip
that around, and you realize that R0 is not an inherent property of the
virus:
“In places with good infection control,
where you can isolate cases as soon as they happen, you’ll see a lower R0 than,
say, in places where an outbreak initially took off.”
Another
key point is an old concept we learnt (and probably forgot) in statistics class
at school:
“R0 is an average. Let’s say the virus has an R0
of 2. This could mean that every single infected person passes the virus to two
other people. It could also mean that one infected person is a “super-spreader”
who infects 100 people, while 49 infected people infect no one. These two
scenarios have radically
different implications for what will happen during an outbreak.”
Yong
isn’t saying that the coronavirus fears are exaggerated (we don’t either way
for sure yet, regardless of the fear). Rather, his point is more nuanced:
“The risk is that a complicated number is
released without context into a world that doesn’t know how to think about it.”
Until this blog, I didn't know such “basic reproduction number”, or just R0 existed. Now I know what it means and its relevance. Well, is the absence of relevance (thanks to this blog)? Well, it is a kind of statistics extrapolation giving it an air of authority, no?
ReplyDelete-----
In the context of statistics application, I recall this exchange between two colleagues when I was at my workplace, EIL. One was a smoker, the other not. The non-smoker was trying to persuade the smoker to refrain from smoking in the office, because some non-smokers complained that smoking by others is a nuisance/suffering for them. Women's voice was prominent in that (maybe those days were different). And, in those days, the law had not been enacted making smoking in public places illegal.
The non-smoker pleaded sincerely the case on behalf of the non-smokers. The smoker laughed all the points away! He was pointing out there was no rule against smoking, which was true. After that, the point that smoking creates disturbance for others is a moral question. Moral questions have a choice between two fates: (1) it can be ignored or rendered irrelevant or ridiculed (2) people's religiosity can be whipped so that mass upheaval will ensure forced acceptance or suffer persecution. In our case, the smoker could have (1) work for him well. :-)
As a last resort, with very little hope, the non-smoker pleaded to the smoker this: The smoker should not be ruining his health with so much smoking. There were many smoking related illnesses, not one or two. Desperately, he mentioned, "See, smoking leads to cancer".
The smoker argued beautifully! He said, "That's all statistical data suggestion. Nobody has proved a cause-effect correlation!" Pretty logical. (Works for engineers!) The non-smoking pleader shrugged his shoulders and walked away. [Incidentally, the smoker died after a complication in prostate surgery long after the incident; he never had any smoking related illness fated for him! So, he did have a point in what he argued, no! :-)] The non-smokers, fortunately, after some time got the relief through a law coming from legislation.
The point I am making is, in the medical domain, many things can only work in the statistical way. Well, we can see that there is "a way of logic" in it. To some extent, statistical correlation is a step suggesting something close to cause-effect way of determination in biological sciences.
Statistical correlation is a tool in physics too, but "almighty laws" can actually emerge from them, unlike any other science. In domains other than physical and biological sciences, possibly statistics is a contrivance, woven to sort of assert anything in one's favor! Every unquestioning lad can be HAD through manipulated statistics, in these domains - I am inclined to believe.