Background: Loneliness is a public health problem that is expected to rise during the COVID-19 pandemic, given the widespread policy of quarantine. The literature is unclear whether loneliness during COVID-19 is similar to those of non-pandemic seasons. This study examined the expression of loneliness on Twitter during COVID-19 pandemic, and identified key areas of loneliness across diverse communities.
Methods: Twitter was searched for feeds that were:(1) in English; (2) posted from May 1, 2020 to July 1, 2020; (3) posted by individual users (not organisations); and (4) contained the words 'loneliness' and 'COVID-19'. A machine-learning approach (Topic Modeling) identified key topics from the Twitter feeds; Hierarchical Modeling identified overarching themes. Variations in the prevalence of the themes were examined over time and across the number of followers of the Twitter users.
Results: 4492 Twitter feeds were included and classified into 3 themes: (1) Community impact of loneliness during COVID-19; (2) Social distancing during COVID-19 and its effects on loneliness; and (3) Mental health effects of loneliness during COVID-19. The 3 themes demonstrated temporal variations. Particularly in Europe, Theme 1 showed a drastic reduction over time, with a corresponding rise in Theme 3. The themes also varied across number of followers. Highly influential users were more likely to talk about Theme 3 and less about Theme 2.
Conclusions: The findings reflect close-to-real-time public sentiments on loneliness during the COVID-19 pandemic and demonstrated the potential usefulness of social media to keep tabs on evolving mental health issues. It also provides inspiration for potential interventions to address novel problems-such as loneliness-during COVID-19 pandemic.
Keywords: COVID-19; Loneliness; Mental health; Natural Language Processing; Social media; Topic modeling; Twitter.
Copyright © 2020. Published by Elsevier Ltd.