Isolate, trace, and test

So remember that time I was trying to understand the contact tracing statistics? Well, the situation has since developed not necessarily to our advantage. According to the most recent release, 68% of suspect contacts are being reached. This is of course less than the official target, a disgrace, etc. However, if the celebrated paper that got all this started is right, even 68% might be easily enough given a R value of 1.5 (it’s currently estimated at 1.1-1.4):

Across different initial numbers of cases, the majority of scenarios with an R0 of 1·5 were controllable with less than 50% of contacts successfully traced. To control the majority of outbreaks, for R0 of 2·5 more than 70% of contacts had to be traced, and for an R0 of 3·5 more than 90% of contacts had to be traced.

Of course, there’s a twist here. The observed R value is partly a function of the contact-tracing effort, and it would undoubtedly be higher if we weren’t tracing. And the effort is real; for those contacts where any contact details were available, the hit rate was over 80 per cent. Somehow, though, things got out of control in September despite an overall tracing success rate that should have been enough to achieve containment even at an initial R of 2.5. What happened? Apart from that thing with the .xls spreadsheet, I mean?

This paper reports the results of a succession of surveys carried out over the summer that tried to measure how thoroughly people were actually self-isolating. The top-line result was not good, suggesting that as few as 18% of people reporting symptoms and 11% of contacts were in fact isolating. There’s an important caveat here, in that they coded anyone self-reporting having any symptoms as being an actual COVID-19 case – this is rather more stringent than the official advice (any of new continuous cough, defined as coughing for an hour, temperature over 37.8 degrees C, loss of smell or taste).

Anyway, the really big, smack in the mouth finding is that the biggest predictors of failing to self-isolate were simply not knowing what the official advice is or what the symptoms were. I don’t really understand the authors’ choices here – they talk a lot about different demographics’ needs, but the effect sizes and p-values on these two factors are huge compared to everything else in the paper. Only half the respondents got the symptoms right, perhaps something to do with the fact anosmia had just been formally declared as a symptom. Respondents who answered questions on the advice correctly were five times as likely to isolate as those who did not.

This is an absolutely crushing verdict on the government’s cacophonous communication, as well as the lack of a serious effort to support self-isolators. After all, the mantra “test, trace, and isolate” is the wrong way around; it’s the isolation that does the actual work.

The problem here, I think, is that individual people and the wider society have different aims regarding testing. When someone goes to the testing clinic, the nurse administering the test hopes to find positives, in order to isolate the infectious and protect the wider society. The patient, though, hopes to test negative, in order to avoid the inconvenience of isolation and to assuage their fear of the virus. Once you understand this, a lot of things fall into place. Why do people do crazy things like trekking across cities to queue at laboratories, coughing? Because they hope a white-coated authority figure, backed by technology, will tell them they don’t have the virus. Why are there never enough tests, no matter how much money we throw at Roche Pharmaceuticals? Because fear is an inexhaustible resource that literally all of us can create, and test reagents have to be manufactured.

The “test, trace, and isolate” model is failing because it is trying to achieve the first set of aims – collective protection – by methods suited to the second, providing tests as a retail product for the worried-well. If I had to propose an answer, I would say that the testing effort should refocus on testing people out of isolation, in order to encourage people to isolate pre-emptively, on suspicion or on warning from contact tracers. The rest of the effort, meanwhile, should refocus on pre-emptive isolation and support to isolators, as well as outbreak investigations. If you expect to be in isolation for a couple of days, you’re more likely to do it than if it’s going to stretch away for weeks. This scheme has the advantage that the testing requirement is considerably smaller, and bounded rather than effectively infinite.

The idea that everyone with a cough in a British winter should lock themselves up for weeks is…I would like to say utopian, but it’s no kind of utopia, and for some reason the word “dystopia” does not convey the same sense of fantasy. Seeing as half the population don’t seem to know the symptoms, we might also consider concentrating on loss of smell or taste, as this seems to be much more specific as an indicator.

2 Comments on "Isolate, trace, and test"

  1. “After all, the mantra “test, trace, and isolate” is the wrong way around; it’s the isolation that does the actual work.”

    Yes, indeed. And at present there are very few policies to facilitate isolation, and the Test and Trace system is not fit for purpose for identifying who should isolate (and who should not). Effective isolation is difficult, so the system ought to identify accurately who should do it (and then help them to do it). That is not happening.


Leave a Reply to Guano Cancel reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.