Impact of Communication Intensity on Vaccination Rate: A Comparative Analysis of Social Media Adoption by the Public Sector of Eastern Europe Countries Using BERTopic Machine Learning Technique

Since the very beginning of the vaccination campaign in Ukraine which started on 23 February 2021 and ended just before the Russian aggression of 24 February 2022 – only 35% of the Ukrainian population had been vaccinated. This is by far one of the lowest results compared with other countries of Eastern Europe (μ=49.73%): Czech Republic (64.2%), Hungary (63.5%), Moldova (40.5%), Poland (59.2%), Romania (42%), Slovakia (50.7%). The only country that has an even lower vaccination rate in comparison with Ukraine is Bulgaria – 29.7%. There is no known reason (like starting date of vaccination, level of trust, variety of available vaccines) to explain the differences in the vaccination rates of these countries. One of the assumptions about the vaccination gap can be given by the fact that both Ukrainian and Bulgarian governmental structures might not have used their social media capacities with full-scale intensity – engaging public sector to communicate crisis information to the population in a convenient, reliable, and comprehensive way. To keep crisis management as effective as possible, any relevant crisis information, especially vital as of Covid-19 vaccination, should be communicated by the governmental channels evenly and synchronized – to increase the overall visibility of key messages and provide a more panoramic view of the situation. Furthermore, coordinated efforts of providing reliable information do reduce a significant amount of “infodemia” characterized by a combination of fakes, fears, rumours, and speculations that strongly affect public sector, media, and the search for the trustful source of information regarding vaccination topic. The question of this study aims to answer is there an impact of public communication on social media platforms such as Facebook, Twitter and Instagram on population behaviour during Covid-19 vaccination campaign. It is also designed to find potential correlations between vaccination rate and the diversity of communication channels, publication frequency, as well as a qualitative aspect of the information given. Given the fact that social media is a part of the general crisis communication scheme, the Pearson’s p-value of our one-year sample taken from November 1, 2020, to October 31, 2021, showed us little to no correlation between vaccination rate and number of messages distributed by the public sector of 8 above mentioned countries. Additionally, we found a low usage of Twitter and Instagram, meaning that two-thirds (8,463 out of 12,911 = 65.5%) of all vaccination messages were concentrated on the Facebook platform. Furthermore, some of the public channels did not use the capacity of their coverage to raise awareness about the vaccination topic at all. We also used BERTopic to cover the qualitative aspect of the data-rich social media study – variation of classical Bidirectional Encoder Representations from Transformers and unsupervised machine learning technique for detecting semantic themes in textual corpora. As a result of using a tool based on a deep learning rather than Latent Dirichlet Allocation’s statistical probabilities, we found some content controversies in Ukrainian, Polish, Czech, Slovak, Hungarian, Moldovan, Romanian and Bulgarian public communication approach towards their population that might also affect vaccination behaviour.