Identifying Novel COVID-19 Rumours Through a Multi-channel Approach

"Tracking rumours is a way to centre communities and encourage a connection between the public health responders and the people they are tasked with serving."

During a health emergency like COVID-19, rumours impacting health prevention behaviours can rapidly emerge and take root. Rumours circulated about COVID-19 vaccination before public officials authorised any vaccines for use, and misinformation intensified as vaccines were rolled out across the world. For many years and across various health areas, Johns Hopkins Center for Communication Programs (CCP), with funding from the United States Agency for International Development (USAID) through the Breakthrough ACTION project, has engaged in identifying and addressing misinformation through strategic communication. This book chapter provides an overview of CCP's rumour identification and analysis process and offers two case studies from the COVID-19 pandemic: working with community-based informants and social media in Côte d'Ivoire, and partnering with a national hotline in Mozambique.

The chapter opens by defining a rumour and describing the process of systematising rumour management, focusing on rumour identification and analysis. In brief, a rumour is an act of communication containing unverified information and can be an event (a case, a death, an outbreak) or a belief (misinformation or disinformation). Rumours are powerful because they resonate with individuals: They can help a community make sense of painful circumstances and regain a sense of control. But they can also create barriers to protective behaviours and undermine the public health response. Therefore, preventing widespread exposure to rumours by rapidly identifying and robustly refuting them is imperative for strong risk communication and community engagement (RCCE).

However, identifying rumours and addressing them pose several challenges, including: (i) Finding the right sources: How does an organisation or government "hear" what people are saying before rumours become prevalent? (ii) Rapidly processing large volumes of unstructured data: How can one identify novel rumours without knowing what one is looking for? Public health communication activities have used a variety of approaches to address health misinformation in different settings, such as pre-bunking, refuting, factual elaboration, myth-busting or testimonials, with varying effectiveness. Whatever the approach, it must be tailored to the local context and take into account social media penetration, trusted influencers, and existing infrastructure.

In order to meet these challenges, CCP in Côte d'Ivoire and Mozambique (as well as several other countries not discussed in this chapter) implemented specific approaches, including:

• Community listening: real-time monitoring and qualitative data analysis. This process involves routinely gathering unstructured comments, feedback, and questions from local liaisons who have been trained to recognise and submit rumours through a mobile application. Community listening also includes more traditional research methods, such as surveys that elicit misinformation or questions that people have. Rumours are entered into a structured form and then coded, using a standard codebook, by data managers.
• Social listening: web and social media analysis to understand online conversations about a topic. Latent Dirichlet Allocation is one approach  that is appropriate for rumour identification because it does not require training data or a priori topics but, rather, groups the data naively, producing a set of topics that can then be evaluated for misinformation.

In Mozambique and Côte d'Ivoire, CCP leveraged existing and new relationships to build a layered network of rumour sources that fed unstructured data to a central database for analysis and action. These rumour trackers, as part of a larger infodemic management system, are cloud-based databases that aggregate rumour submissions for thematic analysis. The case studies that explain this system give a brief overview of the context for rumour management in Mozambique and Côte d'Ivoire, along with any specific nuances of how rumours were identified, and then describe CCP's efforts to address rumours through communication interventions.

1. In Mozambique, the CCP team built on existing efforts by identifying additional rumour sources and creating a DHIS2 database to log, rapidly code, analyse, and visualise rumours. Data were collected from AlôVida, the Ministry of Health (MOH)'s national health hotline; from vaccine acceptance surveys by the United Nations Children's Fund, or UNICEF (one-time online survey) and by Plataforma Educativa de Informação sobre a Saúde (PENSA, 'Health Information Education Platform') (routine ongoing SMS surveys through the text messaging platform from the MOH); from community feedback collected by the Mozambican Red Cross; from social listening conducted by UNICEF; and from input from technical working group (TWG) members. A total of 2,851 rumour submissions were analysed between January 2021 and April 2022. Based on the initial analysis of topics, the TWG agreed on 25 belief statements grouped under 8 main themes, including vaccine deployment, barriers, safety, efficacy, testing and authorisation process, conspiracy theories, prevention measures and modes of transmission. Gradually, the TWG had identified additional belief statements; by April 2022, it had collected a total of 41 unique belief statements.

Next, CCP and members of the TWG coordinated their production and dissemination of communication materials. CCP brought to the TWG its experience in providing technical assistance to the MOH from July 2020 to April 2021. During that period, CCP had developed 12 digital graphics addressing specific rumours on COVID-19 prevention and care, which were published on the MOH website and Facebook page. CCP had also created 14 audio and 9 audio-visual materials broadcast on TV and radio stations at national and community levels. The materials acknowledged the emotions and uncertainty linked to the outbreak and its evolution. Once the vaccine rollout began in early 2021, in line with the initial approach adopted by CCP, TWG members focused on sharing credible, emotionally resonant information and joined efforts to provide testimonials by people recovered from COVID-19 or vaccinated against COVID-19. CCP worked closely with PIRCOM, a USAID-funded Mozambican faith-based non-profit organisation composed of leaders from different religions, to assist in the development of 20 materials showing the support of religious leaders for the COVID-19 vaccine and clarifying rumours. Among other materials the CCP team produced: a set of digital graphic and audio-visual materials using the inoculation approach to foster resistance against fake news about vaccines.

2. In Côte d'Ivoire, at the end of 2019, the Breakthrough ACTION project proposed to the government of Côte d'Ivoire (GoCI) a design for a rumour management platform with real-time visualisation capabilities, which would allow for rapid and concerted decision-making. The implementation began in March 2020, coinciding with the first cases of COVID-19 in the country. In February 2021, with the deployment of the COVID-19 vaccine in Côte d'Ivoire, rumours around COVID-19 vaccination surged. The data in the rumour management system came from community liaisons, hotline operators from health-related hotlines, and radio hosts. Overall, from March 2020 to April 2022, the CCP team, in partnership with the Risk Communication TWG, analysed 6,303 rumour submissions and synthesised them into 83 actionable belief statements.

Next, the CCP team developed 15 posters, 6 video vignettes, and 5 audio vignettes, each in 6 local languages. Each material was focused on a well-defined theme; the selection of themes was oriented by the prevalent concerns found in the rumour data, such as signs and symptoms, barriers, and where to go for testing, with information on official sources of information and contacts for MOH's helplines. These materials were disseminated using a range of strategies, including distribution of posters to community sites and in health facilities, posting and sharing of video vignettes on Facebook and WhatsApp, and broadcasting via traditional mass media channels (radio and television).

Reflecting on these experiences, CCP observes that communication interventions in the 2 settings typically relied on the "factual elaboration" approach, wherein teams offer concise, true statements that respond to - without overly focusing on - the rumours and include a clear call to action.

Identifying rumours through community and social listening, as CCP has done in Côte d'Ivoire and Mozambique, has some limitations. For example, there are challenges with evaluating the approach itself: How do we know rumour tracking works? Does it accurately reflect reality? What populations does it overlook? Does it influence the trajectory of the pandemic?

Despite the need to address such challenges in the future, as argued here, "Tracking and managing rumours systematically as a component of infodemic management allowed public health authorities to hear concerns from communities in their own words."