Monday 27 April 2020

Day 42 of self-isolation - How do we end the lock down?

How do we end the lock down?

First, we need to get the infections and death rates down way below the peak; currently, we are weeks away from there.

Then, experts all seem to agree that we need testing and tracing. Testing to see who is infected, and contact tracing to see who they might have infected.

So, for example. It is discovered that Joe Bloggs is infected. How is this discovered? Either we test large numbers of people (expensive, and way beyond available resources) or else Joe starts coughing, sneezing and has trouble breathing. So he tells his doctor, doctor orders an ambulance, and then we know that Joe is infected.

Now we need to trace his contacts. "Joe, where have you been, and when?" Joe's answer is either "Cough cough sneeze" or else "I really can't remember in detail". "Did you have your phone with you?" "Yes".

So, we can do contact tracing using people's mobile phones. They will have to have GPS services enabled, and be running an app that reports location to a central server once per minute or so. For 63 million people. That's a lot of data. But if we have that data, and can analyse it (that's a lot of computing) we can use Joe's data, and the data of everyone else, to work out who Joe came close to. Then we can contact those people and test them. If any of them are positive, we follow the chain. Trace and test, trace and test.

Before Joe showed symptoms, he was infectious for a week, so we have to check all the people he was near for the whole week. and contact them, and test. Trace and test, trace and test.

It's a huge job. Huge. I can see how the central server program would work, but it'll need a LOT of compute power to do the tracing job. Which I suppose we could rent from Amazon or Google.

But there's another problem. Will people tolerate the huge invasion of privacy that lets the state know exactly where you've been and when? A lot of people won't. Criminals won't. Freedom freaks won't. I certainly wouldn't in normal times, but in the patriotic interest of winning the war against coronavirus, I would. But this isn't going to work unless most people sign up for it. And, of course, there's a ton of people who don't own a smart phone. In the younger age groups, it's 95%, but when you look at the people aged 50 or more, that percentage drops to 50% or so - and those are the more vulnerable people.

And how good is testing going to be? How many false positives, how many false negatives? Each false positive will trigger the test-and-trace chain reaction, as the contacts of that uninfected person are tracked down and tested. And every false negative will be one more chain of infection unsuppressed. If you have  90% accuracy then you have a 10% false positive rate, and Joe Bloggs was close to 100 people over the week he was infectious - then even if all of those 100 people are uninfected, that 10% false alarm rate will give you 10 people you wrongly think are infected, and then you follow the chain of those, and each time you find 100 contacts, that's another 10 people to follow the chain ... this test had better be really good at avoiding false positives.

And if you get a false negative? If half of Joe's contacts were infected, and you find all of them ... except 5 ... then those 5 will start a chain of their own that you don't know about. This test had better be really good at avoiding false negatives.

So is test-and-trace going to work? I don't know, but at the very least, it's going to be an enormous IT project, and governments generally have a very poor record in creating huge IT projects in a very short time frame. And it's going to need a really accurate test system.

80% accurate isn't enough. 

What is the accuracy of Covid-19 tests?

And in America especially, the "freedom first" brigade of unpatriots will insist on their god-given right to do whatever they want, and if that includes spreading the virus, then that's their constitutional right.

So, I'm pessimistic about test-and-trace.

I don't have a better idea, though, so I hope it works.

1 comment:

  1. governments generally have a very poor record in creating huge IT projects in any time frame