Thread by @ProfDFrancis

Suppose, Alice and Bob, two ORBITA HQ PhD students, ask to be put on the same project, so they can spend more time with each other, within the bounds of social distancing rules (and decorum).

Alice takes on the problem of calculating SENSITIVITY of a new Coronavirus test.

Bob, the SPECIFICITY of that test.

Will they be running the new Coronavirus test on the same group of people? (i.e. will they have lots of chances to bump into each other accidentally?)

I used to ask this question on the panel for Academic Clinical Fellowships, before it got banned by the other panelists for being "too annoying".

Every talk on sensitivity and specificity includes a diagram like this:

It makes me laugh because when asked about it, most respondents start off confidently drawing the grid, but then slow down as they try to write things in it.

After some hesitancy, they come up with this.

The key, as an interviewer, is to keep a completely neutral face, giving neither affirmation nor signs of concern.

About half at this point will lose confidence and say they give up.

Another half will make it through to here:

It's amazing how you can go to any amount of lectures on sensitivity and specificity, and nod confidently through them, but walk away with little ability to recall and onwards-teach what you thought you had learned.

There are even mnemonics to add to the jollitude.

Almost everyone has heart of SPIN and SNOUT, although when asked what they mean, there is an audible whirring of mental cogs, and some juddering of the lower lip.

The problem with all these grids, formulae, and mnemonics, is that (for me) I don't find they help. I find they hinder, as they create some barrier of memorisation, and all the understanding is put AFTER that barrier.

Since in day to day life, we are not needing to draw the grid or write the formulae or even interpret the mnemonics, all the "understanding" part does not get rehearsed.

That is why I can easily trick Alice and Bob into working for free for us, in the hope of some romantic reward, that will alas remain unrequited.

Here is my secret method for avoiding all the grids, formulas and mnemonics.

"Mathematics is the art of avoiding arithmetic"

D Francis
Eur Heart J 2020

The key is to realise that I am going to divide my population into two groups, and give Alice and Bob, *opposite* non-overlapping halves!

Alice is studying SENSITIVITY.

This means her subject group is whom?

At 11 votes:

The majority of answers are wrong.

(Thank God) the right answer is just in the lead.

Let me put about half of the audience out of my misery.

In the modern era, this is easy, without a mnemonic or formula.

Now that we are all using video conferencing, we are all used to adjusting stuff on microphones, yes?

What is a SENSITIVE microphone?

What is a SENSITIVE person?

A sensitive test is one which WHEN GIVEN A PATIENT WHO ACTUALLY HAS THE DISEASE, says "Yes, they've got it."

Or,

"Sensitivity - this is your job Alice.

Get people who really do have Coronavirus, and test them. In what proportion *of them* is the new test positive."

It's like saying, "Take this microphone to a dormitory, and find everyone who is snoring. Put it next to them and see in what proportion of SNORERS it is recording a sound."

In other words "Forget about the non-snorers. They don't matter to your PhD"

"By the way, make a list of the non-snorers as you go past them. Bob is gonna go tomorrow and do the microphone test on them."

Here is Bob, on the other hand. He is doing the specificity of the new Covid test.

"Why are you so happy Bob?"

Yes, he had been pretty eager on spending time with Alice, but, being a man, he has his priorities right.

Most important aim in life is to not die.

His role, in measuring specificity, is to test all (and only) people who

You can draw a box and write complicated initials and formulae all you like, but my definition of specificity is just as effective, and is simply this:

SPECIFICITY:

Go get the people who DON'T have the disease.

In what proportion of them is the new test correctly saying that they don't have it?

That's all there is to it.

I will write it in a single tweet so you can memorise it.

Please retweet and pay $1 to Francis Industries any time you use this.

SENSITIVITY =
Get people WITH the disease.
In what proportion is the new test correct?

SPECIFICITY =
Get people WITHOUT the disease.
In what proportion is the new test correct?

D Francis
J It Surely Can't Be This Eas. 2020