CoMix: measuring behavioural change during the COVID-19 pandemic
CoMix is a longitudinal, multi-country social contact survey in representative panels of individuals in terms of age, gender, region of residence and—for most countries—either socio-economic status, occupation or educational attainment. The CoMix study started in March 2020, with survey data first being collected in the United Kingdom (UK), Belgium, and the Netherlands. It was set up to monitor awareness and behavioural changes during the pandemic. Each wave, panel members are invited to fill out the CoMix survey. On the survey day, participants retrospectively report all social contacts made from 5 am on the day preceding the survey up to 5 am on the day of the survey. A contact is defined as an in-person conversation of at least a few words or a skin-contact [9]. For every first wave, the target quota is set at 1500 participants, while a drop-out rate of 5 to 10% is allowed for every subsequent wave (except for some UK panels that were replenished with newly recruited participants and had a higher quota of 2500 later in the survey). When a significant proportion of the panel is lost to follow-up—after sending three reminder invitations—additional panel members are recruited up to the point where the sample matches the target quota. A CoMix wave refers to one period of survey data collection, running from the point when the invitations are sent up to the point when the quota are met and the survey is closed. We refer to the works by Jarvis et al. and Coletti et al. for further methodological details [7, 9]. Apart from participants’ social contacts, the survey also records individuals’ risk perceptions, such as the perceived severity of COVID-19, perceived susceptibility to COVID-19 and the perceived effectiveness of social distancing measures. Due to its longitudinal nature, the survey is particularly suited to quantify how changes in non-pharmaceutical interventions (NPIs) and changes in perception influence NPI compliance and social contact behaviour over time [12,13,14,15]. While most data is collected on behaviour in adults, a proportion of the respondents report contacts on behalf of their children. This provides crucial information about social mixing behaviour in children (and adults) when circumstances change (e.g. schools open versus closed) [8].
In October 2020, several European countries were faced with a surge in COVID-19 cases and had to resort to a second lockdown. Given the diverse range of policy measures in place across Europe [16]—and the central role of social contact data in the parameterization of infectious disease models—the CoMix study was extended to another 17 European countries. In addition, we invited research teams in Norway and Germany (COVIMOD study) that adapted the original CoMix survey to join the initiative in order to set up a collaborative network [17, 18]. The map in Fig. 1 shows the European countries that have been collecting social contact data within the context of—or similar to—the CoMix study. Figure 2 depicts an overview of the available and planned survey waves for all countries. Sample characteristics can be found in more detail in Additional file 1: Tables S1 to S19. We refer to the work by Coletti et al. [9] for sample characteristics for waves 1 to 8 in Belgium.
The CoMix data proved valuable to quantify the impact of social distancing measures over the course of the COVID-19 epidemic in the UK, Belgium and the Netherlands [7, 9, 13, 14, 19], and results are in line with other ongoing studies into social contact patterns with a different study population [19]. The extension of the CoMix study thus provides the opportunity to evaluate policies more accurately within and across a further 17 European countries.
An open-source platform to extract SOcial Contact RATES (SOCRATES) from over 20 European countries
The typical CoMix data flow is reflected in Fig. 3, yet deviations from this scheme are present. The data flow starts from the ‘master’ version of the CoMix questionnaire that was developed and implemented early in the pandemic in the UK (24 March 2020). In collaboration with local partners in each country, the questionnaire is adapted to countries’ circumstances and languages, after which the fieldwork is implemented by a market research company. The CoMix data are cleaned and validated according to a data management protocol, the details and code which can be found on a GitHub repository [20]. After data cleaning, the data is stored and prepared for sharing in the public Zenodo-based repository (accessible via: www.socialcontactdata.org/data) as well as on the CoMix-Socrates tool. Furthermore, we invited partners performing a CoMix-like survey (Fig. 1) to also store their data on the public repository. CoMix data are analysed to gain insights at the national and international level which are converted into advice for health policymakers. The CoMix study protocols and questionnaires were approved—or waivers were obtained—by local ethical committees, the details of which can be found in Additional file 2: Table S20.