Our new preliminary analysis looks at the potential impact of various isolation, contact tracing, testing, and physical distancing measures on COVID-19, using social interaction data from over 40,000 people in the UK cmmid.github.io/topics/covid19… 1/ 
In early Feb, we estimated that additional measures might be needed alongside isolation & contact tracing to control COVID (thelancet.com/journals/langl…). Unfortunately, this would turn out to be the case, from Singapore to Korea. But what are the options now? 2/
In this new analysis, we compared a range of physical distancing and isolation/tracing measures, including app-based tracing, remote working, limits on gathering size, and mass population testing... 3/
We used data from the BBC Pandemic study, a huge citizen science project we ran in 2017/18 in collaboration with @BBCFOUR @FryRsquared & colleagues at University of Cambridge medrxiv.org/content/10.110… 4/
Combining these large-scale data with a transmission model of social interactions and transmission in four settings (home, school, work, other) we were able to simulate where contacts are typically made, and how effective different approaches might be... 5/
We estimated combined isolation/tracing strategies would reduce transmission more than mass testing or self-isolation alone, but still tough to ensure R<1. If combined with physical distancing measures, self-isolation & contact tracing would be more likely to achieve control. 6/
We also estimated around 20-30 contacts would need to be traced per symptomatic case on average. If there were 10,000 new symptomatic cases per day, it meant around 150,000 to 400,000 contacts would be quarantined each day under the scenarios we considered. 7/
Any proposed strategy would of course need to consider logistical as well as epidemiological feasibility, but we outline these numbers to give a sense of what the scale of tracing might look like under different scenarios. 8/
Thanks to the BBC data, we could also categorise contacts into that have been met before and those that hadn't - the latter perhaps being ones contact tracing would be more likely to miss. 79% of work contacts had been met before, but only 52% of contacts in other settings. 9/
This meant combining isolation & contact tracing with restrictions on gathering size outside home/work/school could make tracing relatively more effective, even if some contacts are missed. But size limit may have to be quite small to see effect, potentially <5-10 contacts 10/
App-based tracing could reduce transmission, but because only a proportion of UK has smartphone (and both case and contact would need to have and use app for tracing to work), it would likely need to be combined with other measures. 11/
We also looked Iceland-scale mass population testing (i.e. 0.7% of population per day). Such testing would be very helpful for monitoring the epidemic, but unsurprisingly it had a negligible impact on reducing transmission, because cases would be detected too late (if at all) 12/
In summary: a combination of isolation, tracing and some setting-specific physical distancing may enable control of COVID-19 with less disruption. Designing distancing measures to target hard-to-trace interactions could also enhance relative effectiveness of tracing. 13/
Note again: this is *preliminary work* and has not yet been peer-reviewed. The code is available online (link in the paper at the top), and we're also working on an interactive tool so people can explore other scenarios. 14/
Finally, a huge thank you to the tens of thousands of volunteers who contributed to the BBC Pandemic project. You took the time to help us build a unique dataset on social interactions in the UK, and several COVID-19 projects are now benefitting from these insights. 15/15
Additional note: the 10,000 daily cases above is just an illustrative example (current number likely to be higher in UK currently, because of under-reporting). Key takeaway is that for given number of cases, we estimate 20–30 contacts would need to be traced per case on average.
