Background and objective: Social media could be valuable tools to support people with multiple sclerosis (MS). There is little evidence on the MS-related topics that are discussed on social media, and the sentiment linked to these topics. The objective of this work is to identify the MS-related main topics discussed on Twitter, and the sentiment linked to them.
Methods: Tweets dealing with MS in the English language were extracted. Latent-Dirilecht Allocation (LDA) was used to identify the main topics discussed in these tweets. Iterative inductive process was used to group the tweets into recurrent topics. The sentiment analysis of these tweets was performed using SentiStrength.
Results: LDA' identified topics were grouped into 4 categories, tweets dealing with: related chronic conditions; condition burden; disease-modifying drugs; and awareness-raising. Tweets on condition burden and related chronic conditions were the most negative (p<0.001). A significant lower positive sentiment was found for both tweets dealing with disease-modifying drugs, condition burden, and related chronic conditions (p<0.001). Only tweets on awareness-raising were most positive than the average (p<0.001).
Discussion: The use of both tools to identify the main discussed topics on social media and to analyse the sentiment of these topics, increases the knowledge of the themes that could represent the bigger burden for persons affected with MS. This knowledge can help to improve support and therapeutic approaches addressed to them.
Keywords: Multiple sclerosis; Twitter; natural language processing; sentiment analysis; topic modelling.