The UK’s upcoming contact tracing app aimed at limiting the future spread of coronavirus may not be an effective tool for identify whether users have had close contact with someone carrying the virus, and should not seen as a panacea, according to a study of how Bluetooth signals work in real world situations.
The app, which is being trialled on the Isle of Wight and due for nationwide release later this month, was described this week by the government’s covid-19 recovery document as important to boost “the speed and effectiveness” of coronavirus contact tracing.
However, Doug Leith and Stephen Farrell at Trinity College Dublin concluded it will be “challenging” to correctly record contacts because Bluetooth signal strength varies so much depending on which way phones are facing, whether a body is between two phones and how much nearby materials reflect and absorb signals.
The pair tested four scenarios – walking around streets, at a meeting table in an office, on a train and in a supermarket – using Android phones and a version of Singapore’s tracing app, TraceTogether. Generally, proximity could be established while walking. But at a meeting table, the signal dropped by 38 per cent if both phones were in pockets rather than placed on the table, making it hard to tell if two people had come into close contact.
In supermarkets, it was hard to distinguish between two people correctly social distancing two metres apart and two people walking together at a closer distance. On trains, beyond 3.5 metres, the signal strength between two phones counterintuively increased, against expectations. This may be because the radio waves used in Bluetooth can reflect off the metal surfaces inside trains or in supermarkets, making it difficult to interpret the signals.
Further tests found signal strength dropped by about one-third when someone with a phone in their front pocket faced away from a second person carrying a phone, likely because their body was absorbing some of the signal. Walls may pose a problem too. Although solid cement or concrete walls hugely reduced the Bluetooth signal, stud walls had little impact on signal strength. This means two people in adjacent rooms can erroneously appear to be in close contact, raising concerns over false alarms.
“As far as I know these are the first public measurements from a range of real situations,” says Leith. “They suggest that reliable proximity detection using Bluetooth received signal strength is probably not going to be that easy.”
His work, which has been published online but not yet peer-reviewed, concludes contact tracing apps are “probably not a panacea” but a potentially useful addition to traditional contact tracing.
“It’s not at all clear at the moment what level of errors an app can have and still be useful. Perhaps a high level of errors can be tolerated and the app is still helpful for controlling the spread of infection. We just don’t know,” adds Leith.
Jess Morley at the University of Oxford told New Scientist she would not be installing the app in its current form, because of privacy concerns and fear it would not work well. “If I thought it was genuinely going to work then I would potentially mind less about the other issues [such as privacy] but I think it’s problematically designed and probably likely to be quite ineffective. It’s that combination that makes me say no.”
More on these topics: