More evidence that digital contact tracing worked
The UK is currently the only country in the world that has systematically collected and analyzed app data. A new study reveals just how effective digital contact tracing can be.
The group of Christophe Fraser in Oxford recently published another paper on the epidemiological impact of the NHS COVID-19 app in England in Wales. This is a follow-up to their earlier work published in Nature in 2021, which focused on the first few months of the app during the fall 2020 wave.
The paper by Kendall et al. goes into some detail about how user engagement and the app’s epidemiological impact changed over time. To the best of my knowledge, the UK is the only country in the world that has systematically collected and analyzed app data. Their results provide a unique view into the role that digital contact tracing played during the COVID-19 pandemic, and deserve a broad audience.
App use and engagement
After an initial surge of downloads following the launch of the app in September 2020, app usage remained relatively stable over the following months. In the spring of 2021, as it became mandatory to provide details to the NHS when entering a public venue, the app provided an easy way to do so via QR code scanning. This drove the uptake further from 29% to 38% of the eligible population. Such an increase is quite significant, given that earlier work estimated that each additional percent of uptake reduced cases by about one percent.
Unfortunately, that uptake was almost entirely eliminated by the perceived ‘pingdemic’ in the summer of 2021. The negative connotation around the fact that there were many notifications - when in reality, the app was simply doing its job - was entirely media-driven, and once again points to the importance of good communication.
The percentage of all positive tests that were app-reported reached above 50% in the eligible (i.e., aged 16+) population, which is a remarkable finding by itself, and shows the large public engagement with the app:
Digital and conventional contact tracing
During the first year, there were 2,138,000 app-reported cases and 7,005,000 notifications, leading to an average of 3.28 notifications per app-reported case.
The most remarkable finding in this context is that the conventional (mostly manual) contact tracing showed an anticylic trend, in that it reported reduced numbers of contacts per case during high case loads. At the same time, the digital contact tracing system showed its scalability, performing exactly as expected, with notifications increasing as cases increased:
Testing positive after exposure notification
Contact tracing has an epidemiological impact if it finds infected individuals (preventing further transmission if these individuals isolate). We thus need to understand how effective contact tracing is in finding infected individuals.
The UK app was able to see whether an individual who reported testing positive through the app had recently received an exposure notification. This was the case for a substantial fraction of the app-reported cases:
From this data, one can estimate the proportion of having a positive test after a notification (abbreviated as TPAEN - testing positive after exposure notification). This quantity can’t be directly calculated from the data. Following an exposure notification, many things could happen that prevent an infected contact from entering a positive test result in the app later: the person might delete the app, not take a test, or decide not to enter the positive test result in the app. The paper develops a method to estimate TPAEN, and finds it to be anywhere between 2 - 7%:
TPAEN closely resembles the well-known secondary attack rate (SAR), although it’s not entirely identical. While the previous paper reported an overall estimate over the first few months, the Kendall et al. study developed a method to estimate this quantity over time. Its temporal variability is due to changes in the virus, changes in the immunity landscape (due to infections and vaccines), and other factors.
Similary, the data can be used to compare the daily odds of testing positive for recently notified app users with those of app users who were not recently notified. The results show that the odds ratio was consistently high, and even approached 77 at the peak:
Epidemiological impact
Finally, the paper assesses the epidemiological impact, using an adapted approach of the previously published Nature study. The results suggest that the app prevented about 1 million cases, 44,000 hospitalizations, and 9,600 deaths in its first year. These estimates come with a considerable confidence interval, but are likely underestimates because no indirect effects of app usage were taken into account.
What does it mean
Digital contact tracing was the largest rollout of digital epidemiology technology ever. While many countries launched apps, their public health authorities have not been able to assess the efficacy of their apps because they haven’t collected the necessary data. From personal experience, my hunch is that this was due to a combination of being overwhelmed by the pandemic, and a lack of digital readiness.
Thanks to the data from the UK (or England and Wales, to be precise), we now understand just how effective digital contact tracing was - and how more effective it could have been, had the apps seen more uptake. Many thousands more lives could have been saved. While infection eventually became inevitable for almost everyone, preventing infections before the availability of vaccines was critically important to reduce deaths, hospitalizations, and many of the long-term health consequences.
“Digital contact tracing has played an important role in reducing transmission of SARS-CoV-2 in England and Wales in practice, as was expected in theory.”
In preparing for the next pandemic, we now have a choice to make. Are we going to double down on these low-cost, privacy-preserving technologies that have shown to be both precise and effective, or will we once again resort to medieval methods of quarantining entire populations to keep the virus in check? The answer should be clear. And yet, I see frustratingly little effort being made in serious investments into digital epidemiology. One can only hope that papers like the one by Kendall et al. will manage to push the needle in the right direction.