Photo: Getty Images

Covid-19: The tough questions on a digital contact-tracing app

It’s a crucial factor in keeping the spread of the virus under control, but there’s a whole lot more the government needs to consider than what is technically possible, writes Andrew Chen.

Now we’ve exited level four, how do we keep the Covid-19 cases down? A critical part of the puzzle is contact tracing. In Dr Ayesha Verrall’s report, she notes that rapid contact tracing could be “as effective as any vaccine”. As it stands, contact tracing is largely a manual process: someone gets sick, they get interviewed, and a team of public health officials try to call anyone who might have been a close contact. They let them know they may have been exposed and tell them it’d be best if they self-isolated for a bit. The general idea is that breaking the chain of transmission will keep the spread under control.

But there are well-known challenges with this approach. Interviewing people relies on their memory. Health authorities don’t have up-to-date contact information for everyone. Initially, the phone calls from contact tracers came through to mobile phones as private numbers, so you couldn’t call back if you missed the call, and some people just didn’t pick up because they didn’t know who was calling. This is also a somewhat slow system through which it can take days to reach people and tell them to stay at home. It’s natural to look at this system, and then look at the world around us, and ask “can technology help?”

Unfortunately, it’s not quite as simple as “adding more technology”. We have a lot of options for how we introduce new technology into a system like this – perhaps too many options. And it’s not just about what is technically possible – it’s also about understanding the needs of public health officials, the impacts on human rights, and building public trust and confidence.

Recently, you may have heard about Bluetooth-based contact-tracing apps, with Australia launching its app COVIDSafe late last week. New Zealand might be heading down that track too. So what are the critical questions the government is likely to be considering?

How many people need to use it?

For any digital contact-tracing solution to be effective, people need to be using it. The uptake or adoption rate is the percentage of people in a population who are using the app and contributing data to the system, which indicates what proportion of close contacts we might be able to collect automatically. It’s important to note the government has said that any app is likely to be voluntary (opt-in).

Unfortunately, we’ve never used this technology in a pandemic before, so we don’t have much hard data on what uptake rate is necessary. Estimates vary wildly – some say we need 80% uptake to stop the pandemic, while Australia’s target is just 40% of the population. The oft-cited Oxford paper that models the impact of faster contact tracing suggests even really low uptake rates could help us turn the tide on Covid-19 if we are fast enough.

We also don’t know what sort of uptake rate is achievable. A recent HorizonPoll survey suggests 73% of New Zealanders would support using data from their mobile phone to help with tracking and tracing. But how will this translate into reality? The Trade Me app (one of the most popular apps in New Zealand) sits on about 1.4 million devices, which translates to approximately 28% of the population. Singapore has only achieved around 20% uptake with its app, TraceTogether, and only a proportion of that figure is using it on a daily basis.

However, there is an important piece of contextual framing that changes how we view the uptake rate. In New Zealand, digital contact tracing will only augment or complement the manual contact tracing system, rather than being a standalone replacement. If we are already tracing 60-80% of contacts manually, then even 5-10% on top of that will help, especially if rapid digital contact-tracing methods can help us prevent spread faster.

Usability is also critical for encouraging uptake – we can’t simply say “here’s an app” and expect everyone to download and use it. One of the biggest complaints about TraceTogether and COVIDSafe is that due to privacy and security restrictions on Apple iOS devices, the phone has to be on with the app open on the screen to operate. Having Bluetooth constantly detecting contacts may also hurt the battery life of the device. If it’s not easy to use, people won’t use it! The Apple-Google protocol may help with these usability concerns, as long as people keep their devices updated. But the system also has to be usable for public health officials – if they can’t integrate this with the manual contact-tracing system or it provides conflicting data, then they may ignore the data coming from the digital solution.

How do we balance privacy concerns?

Using an app to collect data about interpersonal contacts does affect individual privacy – the question is whether this is justifiable, necessary and proportional, including whether people will accept it. If people think their data will not be secure or used responsibly, they might avoid participating in the system. Pretty much every proposed app around the world anonymises data, but that’s not enough. The government will have to show that the app is trustworthy, including appropriate legal protections and governance arrangements around who can access the data, what the data can be used for, and when we will shut the system down. This sits alongside suitable security measures such as encryption to ensure that data is safe on a technical level too.

This also feeds into a debate around centralised and decentralised systems, which relates to where data is stored in the system. In brief, centralised systems use apps that send all of their data to a central government server, so that if someone tests positive for Covid-19, it’s easy for health authorities to find that person in the data, as well as all the people they have been in contact with. Decentralised systems keep the data on the phone, so that messages that a person is sick need to be shared through the network of devices, and the devices themselves have to check if they have been in proximity to that person. A purely decentralised system is better for privacy because the government does not get the data, but that would also likely make it harder for the government to integrate it with manual contact tracing. Ultimately, we may end up with a semi-decentralised option where lists of contacts are shared with the government only if someone tests positive.

What if the data is wrong?

The system could make mistakes in two ways. First, the app might think there has been a contact where there hasn’t been (false positives). A criticism of the Bluetooth-based approach is that it may cause neighbours in apartment buildings to be counted as close contacts even if actual interaction is limited by a wall between them. If that happens, people might be unnecessarily quarantined, with the associated cost and inconvenience. In reality, if the app is augmenting manual contact tracing, then human officials could override the app’s recommendations if they don’t make sense.

A second type of error is where the app misses a contact that has happened (false negatives). These are potentially much more dangerous, because a missed contact could go on to spread the disease further. No single system will catch 100% of all contacts, but combining the automated app with manual investigative efforts might do a pretty good job. The government needs to understand what types of contacts will be missed by the app, such as environmental/surface transmission, and try to capture those cases too.

Who misses out?

What about people who don’t have a Bluetooth smartphone, or don’t know how to use it? About 80-90% of people in New Zealand have smartphones, which is quite a high proportion, but still leaves a lot of people out. Certain groups are less likely to have a smartphone or digital device – small children, seniors, people with disabilities, people in rural communities, and people in low socioeconomic communities. These communities may also be disproportionately affected by Covid-19. We can’t have an app for 80% of the population and simply ignore the other 20% – we need to pay special attention to these groups.

This stuff isn’t simple. We won’t get something perfect that ticks all the boxes. We need to be open and transparent about any system’s weaknesses and focus on what might be “good enough”. From what we are hearing in press briefings and through the grapevine, the government is considering these questions and trying to balance it all – these are just difficult questions. InternetNZ has also been doing some excellent work in developing a public conversation around these issues.

The first iteration of “the app” in the next fortnight is likely to be very simple – it will allow people to pre-register and provide updated contact details for manual contact tracers. But remember, contact tracing is just one tool in the epidemiology toolkit – it has to be complemented with ongoing testing, developing better treatments and vaccines, and for our team of five million to keep up good hygiene habits like washing our hands and staying in our bubbles.

Thanks to Tom Barraclough and James Ting-Edwards for their support and contributions to this discussion. Andrew Chen is a research fellow with Koi Tū: The Centre for Informed Futures at the University of Auckland.



The Spinoff is made possible by the generous support of the following organisations.
Please help us by supporting them.