Contacts in context: large-scale setting-specific social mixing matrices from the BBC Pandemic project

Petra Klepac; Adam J Kucharski; Andrew JK Conlan; Stephen Kissler; Maria Tang; Hannah Fry; Julia R Gog

There are other possible biases as well -the day that the participants took part was not randomly assigned. Instead, the participants could choose themselves when to run the app which might have biased some to choose a particularly 'interesting' day when they are going to meet a lot of people, or travel somewhere unusual. In addition, the participants themselves were not sampled at random from the population but instead chose to take part. How they heard about the study might might have varied from whether they were reached through social media in the drive to recruit participants before broadcast, or consequent to watching the BBC4 programme, or through hearing about it from friends -all of which could lead to selection bias. Given the big social media exposure around the citizen science project, our participants were recruited largely in two time periods: in October 2017 after the launch of the app, and in March 2018 after the airing of the documentary 'Contagion! The BBC4 Pandemic', but the uneven recruitment of participants over time should not have much impact. There is evidence, at least among school-aged children, that social contact structure during term-time is relatively consistent over a period of several months [19] . Contacts can also change between term-time and school-holidays [8] and with the health status of participants; individuals typically make fewer social contacts when they have ILI compared to a normal day [28] ]. It may therefore be necessary to combine the matrices presented here with other datasets to fully explore transmission dynamics over long periods and account for changes in behaviour according to health status.