People share and seek information online that reflects a variety of social phenomena, including concerns about health conditions. The study recently published by Alexandre Gori Maia (UNICAMP) and colleagues (Daniel Morales Martínez - UNICAMP, Leticia  Marteleto - UT at Austin, Cristina Guimaraes Rodrigues - FIPE/USP, and Luiz Gustavo Sereneo - UNICAMP) analyzed how the contents of the social network Twitter may provide real-time information to monitor and anticipate policies aimed at controlling or mitigating public health outbreaks. The authors collected tweets on the COVID-19 pandemic with content ranging from safety measures, vaccination, health, to politics. The study then highlights how mentions of selected keywords can significantly explain future COVID-19 cases and deaths in one locality. Two main theoretical mechanisms help explain the links between Twitter contents and COVID-19 diffusion: risk perception and health behavior.

Tens of millions of people search, share, and exchange thoughts and feelings on social media, expressing a variety of reactions that include concerns about health conditions. The study recently published at the Population Research and Policy Review (click here for full access) by Gori Maia and colleagues shows that Twitter data may complement traditional sources of the existing hospital and laboratory-based systems to more accurately predict hotspots of pandemics at the local level and in the short term.

The authors obtained publicly accessible tweets between November and December 2020 containing keywords of interest for the analysis. The study then tested different specifications of spatial econometrics models to relate the frequency of selected keywords with administrative data on COVID-19 cases and deaths one month later. This empirical strategy was able to identify and control for spatial spillover effects of the pandemic and social behavior. For example, tweets from one locality can express feelings about what happens in the neighborhood.

Results demonstrate that mentions associated with social behaviors, in addition to words related to the COVID-19 pandemic itself, can reasonably explain the growth of COVID-19 cases and deaths. The most significant relationships are those with mentions of the pandemic in itself. However, mentions of non-pharmacological interventions, vaccination, health, and even politics are also significantly correlated with future cases and deaths.

The paper finally dicusses how some relationships may be linked to the risk perception of infection. For example, users may express online their concerns over the evolving epidemic and compliance with interventions to prevent the spread of COVID-19. Other relationships may be linked to health behavior. For example, users can reveal their low-risk health perception when expressing opinions regarding health concerns and the benefits of mass vaccination.