Analysis: Modelling on the effectiveness of digital contact tracing solutions shows they are an important tool in our arsenal, but won’t solve the problem on their own, Marc Daalder reports
New modelling from Te Pūnaha Matatini (TPM) analysing the effectiveness of digital contact tracing tools has found that even a solution with massive uptake and considerable effectiveness wouldn’t do enough to reduce transmission of Covid-19 without other interventions.
The paper, which has not yet been peer-reviewed, recommended that digital contact tracing systems be “designed to complement manual contact tracing, for example by enhancing coverage or speed of tracing, rather than as a separate or fully automated system”.
The gold standard, as it were, for a system would be one with 75 percent uptake and recording 90 percent of close contacts.
But even a system with just 40 percent uptake could still be effective.
Bringing down Reff
It all comes down to the reproduction number – the number of people who the average Covid-19 case will infect.
The TPM modelling analyses how the reproduction number might change as different measures are introduced. To begin with, it sits at 2.6 in an environment with no social distancing requirements or other public health measures in place.
The addition of case isolation, in which those who test positive for Covid-19 – on average 6.8 days after symptom onset, based on data from the March/April outbreak – go into self-isolation until they recover, reduces the effective reproduction number (Reff) to 2.4. Manual contact tracing further lowers the Reff to 1.55, when the quarantining of contacts reduces the number of people they are able to infect by 50 percent.
The model assumes manual contact tracers can identify all household contacts, 80 percent of contacts at schools (for children) and half of contacts at workplaces (for adults), as well as just a quarter of casual contacts made while shopping, on public transport, or dining out, for example. Those missing contacts are where digital contact tracing has a role to play.
In the scenario without digital contact tracing or any other interventions beyond case isolation and manual contact tracing, the average outbreak after 30 days would involve 34 cases. The probability of extinguishing the outbreak without other measures would be 67 percent, according to the modelling.
Adding in digital contact tracing with a high uptake rate – in which 75 percent of New Zealanders use the tool – and which logs 90 percent of all close contacts would bring the Reff down to 1.22, the average outbreak size after 30 days to 20 people and the chance of eliminating without other measures to 80 percent. If quarantining of contacts is more effective, reducing the number of people they are able to infect by 80 percent, then the Reff falls to 1.12 and the probability of elimination jumps to 90 percent.
A second tool, recursive contact tracing, could be even more effective. This involves not just tracing and quarantining (through self-isolation) the close contacts of positive cases, but also the close contacts of those first order contacts. With 80 percent effective quarantine, recursive contact tracing and a digital tool which logs 90 percent of contacts and is used by 75 percent of the population, the Reff falls to 1.05 and the probability of elimination hits 95 percent.
The authors warn that recursive contact tracing is difficult to logistically achieve if the outbreak is too large.
“Recursive tracing and effective quarantine of second-order contacts is more difficult to achieve in practice because of the much larger number of second-order contacts and the lower risk of them being infected. In the case on an ongoing outbreak with a large number of cases, the number of uninfected individuals being quarantined under recursive tracing is likely to be prohibitively large,” the wrote.
“However, recursive tracing could potentially be useful in suppressing a small outbreak in its very early stages.”
Digital contact tracing just a tool
All of this sounds impressive, but it still doesn’t reduce the Reff to below 1, which is what is needed to truly slow the rate of transmission, Andrew Chen, an expert in technology and society and a research fellow at the University of Auckland’s Koi Tū – the Centre for Informed Futures, told Newsroom.
“Digital technology is not going to solve everything by itself. It’s just a tool. This centralised, app-based approach with an 80 percent uptake – getting it down to 1.12, that’s still above 1. Even with 90 percent of contacts logged for 80 percent of the population, it’s still only one part of the picture,” he said.
Chen was involved in some of the earlier modelling for this paper but had not seen the final product before it was presented to him by Newsroom.
This is something the paper’s authors acknowledge in their discussion of their results.
“Successful control of Covid-19 is likely to require a range of intervention strategies, including some or all of: moderate population-wide social distancing; widespread use of face coverings; restrictions on large gatherings or other interventions targeting superspreading events; case isolation and household quarantine; manual and digital contact tracing,” the authors write.
“We need to view digital tracing systems as a supporting tool, if you like, for our manual contact tracing systems. They really need to work together, hand-in-hand. And what that means is that when you’re designing a digital system, it really needs to be designed in conjunction with public health units who do the manual contact tracing to make sure that the data these systems provide is actually going to be useful,” Michael Plank, a mathematics professor at the University of Otago and a co-author of the paper, told Newsroom.
However, it can still be very useful. “These results show that digital tracing – if it’s done in a way that supports the manual contact tracing system – can go a long way toward achieving control of an outbreak without resorting to lockdown,” Plank said.
“I think one of the interesting conclusions that the authors make is that, with manual contact tracing you can bring that R effective down to 1.5. And then you’ve got some other sort of physical distancing measures that you have outside of lockdown that might bring it down a little bit more. Then we’ve got mass masking which brings it down a little bit more. So you’re not actually needing a huge impact of digital contact tracing to bring us to a point where we can avoid a lockdown,” Tim Chambers, a senior research fellow in the Health, Environment and Infection Research Unit at the University of Otago, Wellington and an expert on public health and technology, told Newsroom.
Chambers has been involved with the CovidCard project and gave advice on an early draft of the TPM paper.
Certainly, digital contact tracing can never replace the manual process entirely. When TPM modelled this scenario with 50 percent effective quarantine of contacts, a digital contact tracing tool that logged 90 percent of close contacts and an uptake of 80 percent, the Reff was only held to 1.46.
“This could represent a situation where a larger outbreak has exceeded the capacity of the manual contact tracing system, so non-household contacts can only be traced digitally,” the authors wrote. The result “implies that digital contact tracing would need to be combined with significant population-wide control measures in order to avoid a major epidemic”.
The researchers also found that while an uptake of 75 percent was most effective at reducing the Reff, even a tool that is used by two in five New Zealanders would still have some impact.
“It is promising that with only 40 percent [uptake] and recursive contact tracing, that you could see some impact on the epidemic curve,” Chambers said.
Amidst the recent outbreak, that threshold has now been crossed for uptake of the NZ COVID Tracer app. Few digital tools anywhere in the world have achieved anything close to 75 percent uptake.
“I’m glad that we have finally reached 40 percent of the adult population with the NZ COVID Tracer app, but I also hope that the number will continue to increase, especially as Auckland comes out of lockdown and there is more movement of people,” Chen said.
“If the Ministry of Health brings in Bluetooth capabilities [like the Apple/Google exposure notification protocol] into the same app, then we can translate the existing uptake into the new system, acknowledging that some people may not want to use the new system but there may also be others who didn’t see the value in the QR code based system who see value in the Bluetooth approach.”
Which tool is best?
So how would a QR code system fare versus a Bluetooth system that tracks the proximity of one person to another? While the modelling did not reflect the results of particular tools, the authors did attempt to model different potential contact tracing tools by estimating the percentage of close contacts that each solution would log.
A QR code based system was assumed to log just 40 percent of contacts, a Bluetooth app with centralised data control was assumed to log 90 percent of contacts, a Bluetooth app with decentralised control of data was assumed to log 90 percent of contacts but just half of the exposure notifications would result in quarantine because there was no follow-up from the Government and a CovidCard-like device would log 90 percent of contacts, with just 85 percent of those contacts being instantly notified.
“Fully decentralised Bluetooth apps are estimated to be less effective than centralised apps at the same level of uptake, because of a reduced likelihood of users reacting to automatic exposure notifications from a decentralised system without follow up from manual contact tracers. A card-based proximity system is estimated to perform similarly to a centralised Bluetooth app, though with a slightly reduced effectiveness because notifications cannot be sent natively and need to be made via a separate system which requires current contact details,” the authors wrote.
The results of this modelling can be seen in the table below.
However, Chen notes and the authors acknowledge, the modelling cannot fully address other aspects of the systems, including privacy and the alternate uses of a “digital diary” like NZ COVID Tracer.
“The authors note that they didn’t consider the value of having a digital diary and improving speed of contact by helping people get more complete records to contact tracers faster. The Bluetooth methods don’t allow you to keep a diary,” he said.
“They also modelled the QR code exposure notification on an assumption that both parties had to scan the QR code. But as we’ve seen in the last couple of weeks, there is a model where somebody has been in a location, the Ministry of Health interviews them, the case didn’t use the app but the Ministry of Health finds out where it was and then pushes out an exposure notification to people who did use the app and potentially finds contacts through the app.
“There is a little bit of nuance there – but overall, I don’t think that covering those two points would have changed the results by a lot.”
Chambers said the table above shouldn’t be read as endorsing specific options, in part given the fact that each option’s impact is based on assumptions about its effectiveness. Instead the modelling should be used to set up the goal for a digital system, which could be an integrated QR/Bluetooth app or a Bluetooth app that is compatible with a card, which reaches the 75 percent uptake and 90 percent of contacts logged threshold.
“I think we need to stop thinking about it as card versus app and start thinking about those parameters. How can we get the uptake that we need? How can we get the reliability in the technology that we need? So in terms of being able to identify close contacts. And then how do both of those things work with the actual, existing contact tracing system?”
Broadly, the results lay out that digital contact tracing is not a panacea or a silver bullet. But it remains an important tool. With a QR code app with 60 percent uptake or a centralised Bluetooth app with 40 percent uptake, the Reff can be brought to 1.44 or 1.40. What about adding masks on top of that? Or social distancing in public spaces?
The authors take pains to note that reducing the time from symptom onset to test taking – an average of 6.8 days in the model, based on the reality in our first wave – is another lever worth pulling.
“The aim of contact tracing is to find close contacts of confirmed cases and therefore enable early quarantine or isolation. However, it is important to recognise that reducing the time between symptom onset and isolation can bring significant benefits, even in the absence of contact tracing,” they write.
The more tools in our quiver, the less harsh each of them needs to be. Perhaps a more effective app means a better chance of containing outbreaks at Level 2 instead of Level 3. Perhaps masks on buses means it’s okay if just 50 percent of people download the app instead of 60 percent. Perhaps supplementing a Bluetooth app with cards for those without smartphones means gatherings of 100 would be allowed instead of 50.
It’s the combination of measures that, when woven together into a complex fabric, allow New Zealand to pursue elimination of each and every outbreak.