Epidemiological impacts of the NHS COVID-19 app in England and Wales throughout its first year

Abstract

The NHS COVID-19 app was launched in England and Wales in September 2020, with a Bluetooth-based contact tracing functionality designed to reduce transmission of SARS-CoV-2. We show that user engagement and the app's epidemiological impacts varied according to changing social and epidemic characteristics throughout the app's first year. We describe the interaction and complementarity of manual and digital contact tracing approaches. Results of our statistical analyses of anonymised, aggregated app data include that app users who were recently notified were more likely to test positive than app users who were not recently notified, by a factor that varied considerably over time. We estimate that the app's contact tracing function alone averted about 1 million cases (sensitivity analysis 450,000-1,400,000) during its first year, corresponding to 44,000 hospital cases (SA 20,000-60,000) and 9,600 deaths (SA 4600-13,000).

Fig. 1. App usage.

a The number of active app users across England and Wales, and the number of devices with Bluetooth contact tracing enabled. b App uptake per LTLA, estimated as the mean number of active users as a proportion of the total population.

Fig. 2. App engagement.

Weekly numbers of a app-reported cases, b individuals reporting symptoms through the app, and c check-ins via the app’s QR code functionality. Annotations refer to: the steps of a ‘roadmap out of lockdown’; a change to the contact tracing logic of the app for the contacts of asymptomatic cases, and a policy change where some (mainly vaccinated) users were advised to take a PCR test rather than self-isolate upon notification. These events are described in more detail in a timeline in the Supplementary Materials.

Fig. 3. Positive tests reported through the app.

a The percentage of all positive tests by specimen date reported in England and Wales that are app-reported cases. b The estimated percentage of positive tests by specimen date amongst those aged 16 or over in England and Wales that are app-reported cases.

Fig. 4. Notifications and positive tests via digital and manual tracing.

a Weekly rolling averages of app-reported cases and the notifications they trigger, shown on a logarithmic scale. b Weekly averages of contacts notified per index case via the app and via manual tracing (as recorded by CTAS). CTAS contacts are also shown disaggregated by whether they are household or non-household contacts; the household status of some contacts is not recorded so these values do not always sum to the total number of CTAS notifications per index case. The overlap between contacts traced via the app, CTAS, or both, is unknown. Dotted lines represent estimated values. Annotations refer to: the steps of a ‘roadmap out of lockdown’; a change to the contact tracing logic of the app for the contacts of asymptomatic cases, and a policy change where some (mainly vaccinated) users were advised to take a PCR test rather than self-isolate upon notification. These events are described in more detail in a timeline in the Supplementary Materials.

Fig. 5. Testing positive after exposure notification.

a The daily number of app-reported cases which followed an exposure notification, and rolling 7-day average. b The daily and rolling 7-day average percentage of app-reported cases which followed an exposure notification. c The mean estimated proportion of notified individuals who enter a positive test result into the app shortly after exposure notification (TPAEN). Shading around the line indicates the 95% credible interval. d The probability of testing positive after notification relative to a random member of the 16+ population, shown on a logarithmic scale. The central estimate and shading around the line correspond to the uncertainty reported for the ONS infection survey (official estimate and 95% credible interval respectively). The horizontal dotted line indicates a value of 1, i.e., equal probability. e Daily new app-reported cases per 100,000 active users by notification status. f Daily maximum likelihood estimates of the odds for testing positive in recently notified app users relative to not-recently-notified app users. Shading indicates 95% confidence intervals. The horizontal dotted line indicates a value of 1, i.e., equal probability.

Fig. 6. Epidemiological impacts.

Cumulative estimated numbers of a cases, b hospitalisations and c deaths averted by app exposure notifications between 24 September 2020 and 24 September 2021. Shading in panels a–c indicates the range of outcomes between upper and lower plausible estimates of an individual’s reduction in risky contacts as a result of receiving an app notification, while the central estimates correspond to moderate reductions in risky contacts. d Estimated cases averted in each LTLA. e Estimated percent reduction in cases in each LTLA.