Dynamic Topic Modeling of Kratom Use and Experiences: Insights on 13 years of Reddit Discussions

Authors Seraphina Fong, Alessandro Carollo, Elisabeth Prevete, Ornella Corazza, Gianluca Esposito Abstract Traditionally used in Southeast Asia for common ailments, kratom (Mitragyna speciosa) is increasingly being adopted in other regions for the self-treatment of pain, mental health issues such as anxiety, depression, and dependence or managing the withdrawal symptoms linked to stimulants/sedatives or other substances in absence of clinical supervision. To gain better insight into the experiences and the motivations underlying kratom use, previous studies have analyzed material shared among users on social media. However, these investigations often employed manual labeling and categorizations, which are time-consuming and do not allow the investigation of large samples. The present work utilized a natural language processing approach (BERTopic) to extract prominent topics of discussion from kratom-related Reddit posts made between 2010 to 2023 (n = 188,139 posts). Dynamic Topic Modeling was also implemented to analyze how discussions about these topics evolved over time. From the results, users discussed topics including usage for substance use disorder recovery (e.g., suboxone), drug testing, consumption methods, logistic issues (e.g., bringing kratom on flights, receiving kratom via mail), tapering off the use of kratom, and providing support to others who are in the process of quitting kratom. The current study supports the need of clinical trials as well as new ecological insights into experiences of kratom users, while supporting the implementation of new policies to regulate its usage. Public Data https://www.reddit.com/r/pushshift/comments/1akrhg3/separate_dump_files_for_the_top_40k_subreddits/ Public Preregistration The author asserts that a preregistration is not applicable because no data collection, extraction, or analysis is reported in the preprint. Conflict of Interest Author asserted no Conflict of Interest. Supplemental Materials https://osf.io/24ajk preprint DOI https://doi.org/10.31219/osf.io/5hyjx_v3