Can you trace contacts with a smartphone?

In my previous article, I concluded that the primary purpose of the new Covid-19 smartphone app being trialled on the Isle of Wight was early reporting and testing of suspected cases, rather than tracing contacts. That doesn’t imply that the latter is unimportant, but it makes a centralised reporting model essential, and explains why the decision was made not to use Google and Apple’s new frameworks – which aren’t available yet in any case.

This article looks at the benefits the app might bring, both in terms of case detection and contact tracing.

Diagnose and test

Around 50,000 of the Isle of Wight’s population of about 140,000 have downloaded and installed the new app. That’s just over 35% of the whole population. Inevitably, some of those users will never open the app, or will remove it again. Assuming those numbers are relatively small, that means that 35% of the population are now likely to report their initial symptoms should they suffer overt Covid-19. For public health monitoring, that’s an excellent proportion.

As no one knows how many cases have already occurred on the Island, because testing has been largely confined to those who were sick enough to be admitted to hospital, it’s hard to predict how many cases might self-report using the app. If about 20% of all Covid-19 cases have been admitted to hospital, and the rate of occurrence on the Island remains unchanged, there might be 10-20 fresh cases each day in the whole population.

That would mean between 3-7 cases per day would self-report using the app. If daily numbers were to rise above 10, then that might be sufficient to ask whether an outbreak was developing. Initial results come in provisionally on the same day as those first symptoms, with test confirmation following 2-3 days later. In infectious disease control, you can’t get much better.

Early reports from NHS England claim that as many as 25 fresh cases are reporting through the app each day. Not all of these will prove to have Covid-19, though. There is strong evidence that Covid-19 has been greatly under-reported, and this demonstrates how central reporting can provide public health data of major value in managing the infection and its prevention.

Contact tracing

The purpose of tracing contacts is to limit further spread, so that each fresh case of Covid-19 infects less than one additional person. When that happens, numbers of fresh cases will fall and, given sufficient time, the infection will eventually die out of its own accord (although in practice life isn’t as simple).

Assume a simple model of someone who, following relaxation of lockdown, travels by public transport to work, works a normal shift, then returns home by public transport. During each journey, they come into close contact with two other people; at work they come close to a further two, giving a total of six close contacts in the day. The following morning, they feel rotten, and have the early symptoms of Covid-19.

If no effort is made to trace their contacts, six people have been exposed to that source of infection. If there’s a 50% chance that each will have caught Covid-19, then (everything being truly random) on average that single pre-symptomatic case is likely to have infected three others. In practice, in the period immediately before the lockdown took effect on 21 March, that’s what was happening with Covid-19 across the UK.

Using traditional methods, a good contact tracer would be able to identify and warn that person’s two colleagues at work. But they would be unable to discover who travelled on the same public transport. Even with early testing and quarantine, two fresh cases would still be travelling around, in their turn spreading Covid-19.

The purpose of smartphone-based contact tracing is to identify and quarantine those four anonymous contacts, who aren’t accessible using traditional methods. If the new app can contact half of them, that leaves just two unaccounted for, of whom (on average) only one will go on to develop Covid-19: instead of infecting three or two others, the index case would infect just one, and fresh case numbers would remain stable rather than rising again.

Smartphone contact tracing

More sophisticated modelling is needed to determine the percentage of the active population required to make smartphone-based contact tracing yield sufficient traces to ensure that Covid-19 infection rates don’t increase. However, it’s simple to gain insights using basic probabilities.

In this case, the percentage of users works in our favour: for contact tracing, we’re not concerned about covering the whole population. In the case of the Isle of Wight, it’s unlikely to be a total of any more than 100,000, giving a current app adoption rate of 50% or better. Using that as a starting point, and assuming that the app works perfectly for every close contact provided both people have it installed, the chances of each potentially infecting contact being registered as a contact are 0.5 x 0.5 = 25%. In the other 75% of contacts, one or both of them won’t have a smartphone with the app installed, and will not register the contact.

Using the example figures above, the new app would only pick up one of the four anonymous contacts, leaving three, of whom on average 1.5 would develop Covid-19, leading to a slow growth in fresh cases.

In this simple model, it would take 75% of the active population to be using the app at the time of each contact, to detect and trace half the anonymous contacts and keep the infection rate stable.

On the Island, that would require approximately 75,000 of the population to use the app, just over half the total population. That seems unlikely there, although such high adoption rates could be achievable in some urban areas, where anonymous contacts are likely to be more frequent.

The conclusion of studies such as that by Kucharski and others, which use sophisticated mathematical models, is that smartphone contact tracing is going to have to be combined with distancing measures if the infection rate is to be controlled. Clearly it is only part of the solution, and early diagnosis and testing are crucial.

This all assumes that, when two smartphones with the app installed do come into contact, that is successfully registered. One of the many claims by critics of the new NHS app is that isn’t possible for iOS. In the next article about the app, I’ll consider how reliable it is at registering contacts, and what insight it gives into what is considered to be a high-risk contact, something no one seems to be prepared to tell us.