TLDR: although a model has suggested mass-testing to be ineffective in controlling COVID-19, if we adjust the parameter for the scale of testing, and combine the approach with contact tracing, the model predicts that R can be reduced below 1 without social distancing.

This is a fork of https://github.com/adamkucharski/2020-cov-tracing which is the code that contains the analysis performed in Effectiveness of isolation, testing, contact tracing and physical distancing on reducing transmission of SARS-CoV-2 in different settings by Adam J Kucharski, Petra Klepac, Andrew Conlan, Stephen M Kissler, Maria Tang, Hannah Fry, Julia Gog, John Edmunds and the CMMID COVID-19 Working Group.

As part of their paper, the authors investigate many measures for controlling COVID transmission, including mass-testing. A reader of the paper might conclude that mass-testing has little to offer in terms of controlling COVID, since the paper argues that R_eff is 2.6 without any control measures, and almost the same (2.5) with mass-testing.

The paper models mass-testing only in the following scenario:

• 5% of the population is tested per week
• when positive cases are identified by mass-testing no attempt is made to isolate or trace their contacts

The paper concludes that mass-testing can contribute little. The only scenario modelled that offers an R_eff below 1 requires manual contact tracing, isolation of symptomatic people, and substantial social distancing (only 4 contacts per day allowed outside of school/work). This achieves an R_eff of around 0.93.

Here we have made two small changes to the code to enhance the effectiveness of mass-testing. Firstly we substantially increase the proportion of the population that can be tested each week, from 5% to 37%. This is based on a proposal for mass-decentralised testing in schools and workplaces.

Secondly, we attempt to add isolation and manual contact-tracing to the model alongside the mass-testing intervention. It seems perverse to perform mass-testing and then fail to trace the contacts of people identified through it, many of whom will be in the pre-infectious incubation phase.

We are not convinced we have adequately modelled the full effect of quarantining the contacts of anyone who tests positive under mass-testing (see issue #1). Nevertheless, the two changes made here result in an R_eff of 0.92, lower than any of the other scenarios and without the need for social distancing.

If we are more optimistic about the ability to perform mass-testing, and we are prepared to pair it with contact tracing and isolation, then mass-testing may have a role to play in controlling COVID-19, while enabling life to largely proceed as normal. If we are considering whether to build capacity for mass testing, we should evaluate the possible reduction in R for whatever capacity we are considering, and should acknowledge that mass-testing will always work much better when paired with contact tracing.

99.999% of this code is by the original authors. This re-analysis is only possible because they have made their code available.

This has been put together by a non-expert. Corrections are welcome.