COVID-19: will contact tracing work?
It’s widely said that testing and contact tracing (using a smartphone app) is the key to the control of COVID-19. But it is not necessarily that simple; there are several unknowns that are worth thinking about.
The textbook example that underpins much of the thinking about this issue is the eradication of smallpox. The later stages of that campaign relied on the early detection of cases and the vaccination of “contacts” (in this case, everyone within a certain area). There are several factors underlying the success of that strategy.
The most obvious is the availability of an effective vaccine. We don’t have that (yet?) for COVID-19.
Secondly, the symptoms of smallpox were obvious and distinctive. There was no need for a complex and time-consuming test to identify a case. In the final stage, in remote areas of Ethiopia, one person in each village was trained to spot, and report, cases, so the vaccination team could respond quickly.
A third important factor is that smallpox is not infective until symptoms appear. Here there is considerable uncertainty in the comparison with COVID-19. There is circumstantial evidence that transmission may occur from pre-symptomatic individuals (i.e., those who subsequently develop symptoms) and possibly also from asymptomatic people (who never develop recognisable symptoms) – although probably to a lesser extent than from those who have symptoms. But we don’t really know, and if it happens to a significant extent it could reduce the effectiveness of a contact tracing strategy. On the other hand, if R is less than one for pre-symptomatic/asymptomatic individuals, then it might not matter.
The comparison with smallpox does have one favourable factor. Both diseases have a relatively low R0 value (2-4)*. This is a marked contrast with another textbook example – measles, where the R0 value is much higher (15-20 is often quoted). Measles also provides another contrast, in that the initial disease is an inconspicuous respiratory tract infection. This is the infectious stage. The typical symptoms come later (they are an immunological response to the virus), and by that stage the patient is not infectious. The high R0 value would make a contact tracing strategy extremely difficult; infectivity before detection makes it effectively impossible.
*A digression to clear up a point that often causes confusion. For epidemics in general, R0 (the basic reproduction rate) is the value of R at the start of a new epidemic, when everyone is susceptible. As the epidemic progresses, and the number who have had the disease (and become immune) increases, the value of R (the actual reproduction rate) declines. For COVID-19, we are largely looking at the effect of control measures rather than the number of those who have become immune; nevertheless it is simpler to refer to this as an effect on R rather than R0.
Before getting back to the point – will contact tracing work in enabling a relaxation of the lockdown while still keeping R below 1? – we need a more subtle interpretation of R. It is an average value across the whole population. If R is very low for some people and much higher for others, you could still get an average value of R <1 even if there is a subpopulation that is spreading the virus quite effectively. If this is geographical (a rural-urban distinction for example) it will show up quite readily (and already does). But if it applies to different groups within say a major city, it is not so easy to see.
Furthermore, although a value of R<1 is (rightly) regarded as a significant point in predicting a decline in the epidemic, it is not an absolute objective. The further it can be reduced, the quicker the epidemic will die out (as well as countering the possibility of a sub-population of spreaders.
Now we can think more clearly about contact tracing. Firstly it depends on the relationship between infectivity and symptoms. Assuming, for the moment, we will not be undertaking massive random testing of the whole population, the contact tracing app will largely apply only to those who are symptomatic. We would have to hope that for those without symptoms, R is already quite low – i.e., there may be a possibility of transmission, but at a low frequency.
Secondly, there is the time factor. Once someone develops symptoms, how soon would they become aware of that and report it? One day? Two days? Then the app needs to notify the identified contacts, who are expected to self-isolate. The crucial factor here is whether those contacts will self-isolate quickly enough, before any of them have become infectious. If the original person is infectious for two days before reporting it, and if the identified contacts don’t act quickly, they may well have been infectious themselves for a day or so before they self-isolate.
The third factor is the degree of uptake of the app. The government originally predicted 80% but have backtracked on that to a figure of 50%. Many people regard that as optimistic – some predictions put it as low as 20%. (Bear in mind that only 79% of adults have a smartphone; for those over 65, it is 40%). Concerns over privacy could have a major impact – I can think of several examples of people who wouldn’t want to disclose who they have been in contact with, however much they are told the data is anonymised!
Then we come to the question of numbers. Suppose we have 10,000 cases per day. (Yesterday, 6 May, there were just over 6k positive tests in the UK, from 57,000 people tested, so the actual number of cases is probably much higher than this). And suppose each case had 5 contacts. That makes 50,000 people per day told to self-isolate for a week – giving a total of 350,000 people self-isolating at any one time. The vast majority will not develop symptoms and may regard the exercise as a waste of time. Is this sustainable?
Of course, if distancing works, there shouldn’t be any contacts, but a lot depends on what the app considers as a ‘contact’. How close do you have to be, and for how long?
You may notice that I’ve said very little about testing. Despite all the publicity about the number of tests being done, and how important it is to do much more, it is far from clear how this would contribute towards a contact tracing strategy. If the identified contacts were tested, would this discriminate between those who had been infected and those who hadn’t? If it did, you could release some from isolation. But there is some doubt as to how early in infection the test result becomes positive, and with the current delays in getting the result back to the subject, they would be half-way through their isolation before they were told the result.
Or, if we did random testing of the population, that would identify a lot of people who were infected but asymptomatic. They could be told to self-isolate, and report it to the app, resulting in a large increase in the number of contacts who would in turn need to self-isolate. But if someone who is asymptomatic transmits the disease only occasionally, this does not have much of an effect.
It would be nice if those advocating a large increase in testing would be more explicit about why it is important. I should emphasise that I’m thinking in terms of control strategy in the general population, and not staff and patients in hospitals and care homes, and similar situations, which is a very different matter, and the need for testing there is quite clear.
The final assessment is that potentially it could work, but it will require a very effective publicity campaign, first to convince people to use the app, and then to take notice of the requirement to self-isolate if they have had a reported contact.