Examining impact of Twitter sentiment & source on COVID-19 info sharing by public

INTRODUCTION: Public health communication programs have the power to affect change in people and groups by increasing awareness, enhancing knowledge, modifying attitudes and altering behaviours when they are properly designed, meticulously executed and maintained over time. There has been an influx of public health communication on social media during COVID-19. AIM: To investigate if the Twitter user category (source) and/or message type (sentiment) impacts sharing behaviour of the public during COVID which would help understand how to use social media when communicating health messages. METHODS: All tweets in English were extracted using Twitter API and Python script. Data was extracted from 1 Jan 2020 to 31 Mar 2022 for Ireland and UK which resulted in 867,485 tweets in total. Four coders identified user categories by studying the user descriptions in an iterative process and sentiment analysis was conducted to identify message types during the pandemic. For sentiment categorisation, a similar approach was followed where four message categories were identified from existing public health literature and additional categories were created through the analysis of Twitter messages by five coders. Different Machine learning (ML) models and Chat GPT were applied to assign user categories and sentiments to the entire dataset and the model with the highest level of agreement with human coders was chosen. RESULTS: Overall the public sentiment was negative during the pandemic. Twelve user categories and seven message types were identified from our analysis. CONCLUSIONS: Further detail of this ongoing analysis will be presented at the conference. KEY MESSAGES: • The abstract introduces a study that investigates the impact of Twitter user categories on the sharing behaviour of the public during COVID-19. • The study investigates the impact of message types on the sharing behaviour of the public during COVID-19 on Twitter.