Interesting read from @ruskin147. How can we know if contact tracing apps are working? A thread. 1/n
First, the theory. We know contact tracing is effective against the spread of infectious diseases. It has been used for centuries in combination with isolation of infected people, and is credited with suppressing many epidemics including SARS-COV-1. 2/n
The current coronavirus pandemic, SARS-COV-2, is much harder to suppress with contact tracing and self-isolation because it spreads so quickly and people can be infectious without symptoms. That's why contact tracing apps were proposed. 3/n
We have the theory: digital contact tracing could work quickly enough and at large enough scale to suppress the pandemic. 4/n
doi.org/10.1126/scienc…
doi.org/10.1126/scienc…
Now we need to know the practice: many countries have released such apps - are they having a positive effect and reducing the spread of the virus? 5/n
A brief aside: the article says "it is thought more than half of the population need to have an app before it is truly effective". This is a bit of a misunderstanding and I see it a lot so allow me to clarify. The result concerning ~60% of the population is: 6/n
If about 60% of the population used a [well-calibrated, perfect] contact tracing app then *that alone* could suppress the epidemic - you wouldn't need social distancing, masks, etc. Which is phenomenal.
NO-ONE is advocating we do this. 7/n
NO-ONE is advocating we do this. 7/n
Contact tracing apps are new technology, they won't always work perfectly, people won't always follow their guidance. Moreover, they need to be integrated into a robust public health system in conjunction with human contact tracers and fast testing strategies. 8/n
Contact tracing apps should be used alongside fast testing, human contact tracing, social distancing and mask wearing. In that setting, the theory says they will reduce the spread of the virus *even with low uptake*. Let's not get hung up on the 60% thing. 9/n
Back to the practice. Is there any evidence for contact tracing apps working? In short: not yet. Early results from the Isle of Wight point towards success for the *combined* Test and Trace strategy there but we need more data to confirm. 10/n ox.ac.uk/news/science-b…
What sort of data would help to evaluate these apps? Especially the more privacy-preserving ones? Cellan-Jones and Kelion say
"The problem is that with so little data exposed, can we even distinguish good from bad?" 11/n
"The problem is that with so little data exposed, can we even distinguish good from bad?" 11/n
One thing not mentioned in the article which we could be measuring is:
*For each person who takes a test, have they recently been (human/app) contact-traced?*
That's a strong indicator for how well tracing is staying one step ahead of the virus. 12/n
*For each person who takes a test, have they recently been (human/app) contact-traced?*
That's a strong indicator for how well tracing is staying one step ahead of the virus. 12/n
It helps towards establishing the four crucial measures:
1) the number of people who were traced who were infected
2) the number who were traced who were not infected
3) the number who were not traced who were infected
4) the number who were not traced who were not infected
13/n
1) the number of people who were traced who were infected
2) the number who were traced who were not infected
3) the number who were not traced who were infected
4) the number who were not traced who were not infected
13/n
All of these are important. A perfect system would have high proportions of (1) and (4) and low (2) and (3).
HOWEVER.
There's a subtlety about (2) which is worth a moment's thought. 14/n
HOWEVER.
There's a subtlety about (2) which is worth a moment's thought. 14/n
Suppose I was contact-traced, I self-isolated for 14 days, didn't develop symptoms and tested negative. We could describe this as a failure of contact tracing, a major and unnecessary inconvenience. And it sort of is. BUT. 15/n
What if I genuinely had been "riskily" close to an infected person? Even a perfectly functioning app / human system will identify times I got close but, happily, avoided becoming infected. My self-isolation might in hindsight appear unnecessary because I wasn't infected. 16/n
But what would have happened if I hadn't self-isolated? I had been near an infected person in the workplace / on my bus route / in the shop, so there is a non-negligible probability that they infected someone else in my workplace / on my bus route / in the shop. 17/n
Being contact traced is primarily intended to avoid an infected person spreading their infection to others. But by self-isolating at that time, I might actually have stayed home at just the right time to *avoid getting infected at all*. 18/n
What my contact tracing notification in fact tells me is
"coronavirus is circulating near you. If you're prepared to spend 14 days in self-isolation, now is a really good time to do it."
And that's a very powerful intervention. 19/n
"coronavirus is circulating near you. If you're prepared to spend 14 days in self-isolation, now is a really good time to do it."
And that's a very powerful intervention. 19/n
Evaluation of (human/app) contact tracing will be in the news for a while. It will be important to remember the nuance about
"2) the number who were traced who were not infected"
because it might just be that the contact tracing itself saved them from being infected.
20/20
"2) the number who were traced who were not infected"
because it might just be that the contact tracing itself saved them from being infected.
20/20
