Best Paper Selection

Appendix: Content Summaries of Selected Best Papers for the 2024 IMIA Yearbook, Section Consumer Health Informatics

Niederkrotenthaler T, Tran US, Baginski H, Sinyor M, Strauss MJ, Sumner SA, Voracek M, Till B, Murphy S, Gonzalez F, Gould M, Garcia D, Draper J, Metzler H.

Association of 7 million+ tweets featuring suicide-related content with daily calls to the Suicide Prevention Lifeline and with suicides, United States, 2016-2018.

Aust N Z J Psychiatry. 2023 Jul;57(7):994-1003.

doi: 10.1177/00048674221126649.

The study examines the relationship between the content of tweets related to suicide and their impact on calls to the US National Suicide Prevention Lifeline (Lifeline) and actual suicides. The study aims to understand how different types of social media content influence help-seeking behavior and suicide rates, utilizing machine learning to categorize tweets and perform time-series analysis. The study retrieved 7,150,610 tweets geolocated to the United States, posted between January 1, 2016, and December 31, 2018. Tweets were categorized into six content areas using a machine-learning approach: prevention, suicide awareness, personal suicidal ideation without coping, personal coping and recovery, suicide cases, and other. Seasonal Autoregressive Integrated Moving-Average (SARIMA) models were used to assess the associations between tweet categories and daily calls to the Lifeline and suicides. The analysis focused on the immediate impact (same-day associations) of tweet content on help-seeking and suicides, given the short lifespan of attention to tweets. The analysis of tweet content revealed that 15.4% of the tweets were awareness tweets, aimed at spreading awareness about the problem of suicide. Prevention tweets, which focused on sharing information about how to prevent suicide, accounted for 13.8% of the tweets. Tweets discussing specific suicide cases made up 12.3% of the total. Tweets that mentioned suicidal ideation without any indication of coping represented 2.4%, while tweets that shared personal stories about coping and recovery from suicidal thoughts were the least frequent, at 0.8%. A significant portion of the tweets, 55.4%, were categorized as other or irrelevant. The study found that prevention tweets were positively associated with calls to the Lifeline, with an estimated coefficient (B) of 1.94, a standard error (SE) of 0.73, and a p-value of 0.008. Similarly, coping tweets also showed a positive association with Lifeline calls, with an estimated coefficient of 9.40, a standard error of 3.53, and a p-value of 0.008. Conversely, tweets about suicidal ideation or attempts without coping were negatively associated with Lifeline calls, with an estimated coefficient of -5.79, a standard error of 2.52, and a p-value of 0.02. Additionally, the total number of tweets was found to be negatively associated with Lifeline calls, with an estimated coefficient of -0.01, a standard error of 0.0003, and a p-value of 0.007. The analysis also revealed that prevention tweets were negatively associated with the number of suicides, with an estimated coefficient of -0.11, a standard error of 0.05, and a p-value of 0.038. No significant associations were found between tweets about suicide cases or coping stories and the number of suicides. However, the total number of tweets was positively associated with suicides, with an estimated coefficient of 6.4x10-5, a standard error of 2.6x10-5, and a p-value of 0.015. The study is the first large-scale analysis to suggest that specific suicide-prevention-related social media content on Twitter corresponds to higher levels of help-seeking behavior and lower numbers of suicide deaths. Prevention tweets, which often included information on resources like the Lifeline, were associated with more Lifeline calls and fewer suicides. Coping tweets, though less frequent, also showed a positive impact on help-seeking behavior. Conversely, tweets expressing suicidal ideation without coping mechanisms were linked to fewer Lifeline calls. The results support the idea that the content of social media posts can significantly impact public health outcomes. Prevention and coping tweets may encourage help-seeking and reduce suicide rates, highlighting the importance of promoting positive and supportive content on social media.

The article directly addresses the topic “Precision in Prevention and Health Surveillance: How AI May Improve the Timeliness of Identification of Health Concerns through Social Media Content Analysis” by demonstrating how machine learning can categorize large volumes of social media data to provide timely insights into public health issues. The use of AI to analyze social media content allows for real-time monitoring and identification of health concerns, facilitating immediate interventions. The study showcases the potential of AI in enhancing the precision and timeliness of health surveillance efforts, particularly in the context of suicide prevention. By utilizing advanced AI techniques to process and categorize tweets, the research highlights how AI can improve the detection and response to health-related issues through the analysis of social media content, ultimately contributing to better public health outcomes.

Smith BP, Hoots B, DePadilla L, Roehler DR, Holland KM, Bowen DA, Sumner SA.

Using Transformer-Based Topic Modeling to Examine Discussions of Delta-8 Tetrahydrocannabinol: Content Analysis.

J Med Internet Res. 2023 Dec 21;25:e49469.

doi: 10.2196/49469.

This article investigates the public health implications of delta-8 THC, a psychoactive cannabinoid increasingly available due to hemp legislation. Given the delta-8 THC potential health risks and the surge in use, the study aims to understand the discussions surrounding delta-8 THC on social media, specifically Reddit, using advanced natural language processing (NLP) techniques to identify trends and health-related concerns. The study analyzed posts from 115 cannabis-related subreddits on Reddit, collected from January 2008 to December 2021. Data was retrieved using the Pushshift.io API, focusing on posts mentioning delta-8 THC. The researchers employed BERTopic, an unsupervised topic modeling approach based on transformer models, to identify and cluster discussion topics. Preprocessing included minimal text cleaning and the use of Word2Vec to identify various terms for delta-8 THC. Topics were validated by human experts to ensure accuracy. The study used publicly available data and ensured the anonymity of Reddit users. It was exempt from institutional review board oversight by the CDC. The study identified 41,191 posts related to delta-8 THC out of 44,719,379 cannabis-related posts. From these, 81 distinct topics were identified, with the main topics being: 1) Product Discussion: 27.6% of posts discussed specific delta-8 THC products like edibles and vapes; 2) Comparison to Other Cannabinoids: 19.8% of posts compared delta-8 THC to other cannabinoids like delta-9 THC and CBD; 3) Safety Warnings: 13.8% of posts were warnings, often generated by bots, about product safety and misinformation; 4) Legality: 11.2% of posts discussed the legal status of delta-8 THC; 5) General Interest: 6.1% of posts expressed curiosity or general experiences with delta-8 THC; 6) Health Concerns: 5.4% of posts discussed health effects, both therapeutic and adverse, such as anxiety relief and breathing issues. Mentions of delta-8 THC increased significantly from 2019 to 2021, suggesting a growing interest and use. Health-related posts indicated both positive effects, such as anxiety relief, and negative effects, like respiratory issues. The study highlights the rapid increase in delta-8 THC discussions, reflecting its growing popularity and potential public health impact. The diversity in product forms and consumption methods suggests widespread use and accessibility. Legal discussions indicate confusion and concern about the regulatory status of delta-8 THC, driven by its classification under the 2018 Farm Bill. Health-related posts reveal mixed experiences, with some users reporting benefits and others noting adverse effects, underscoring the need for further research and regulation. The use of BERTopic provided detailed insights into the diverse topics of discussion surrounding delta-8 THC, outperforming traditional topic modeling approaches. Limitations include the lack of geographic and demographic data, potential non-representativeness of Reddit users, and challenges in categorizing unstructured text. The study demonstrates the utility of social media data in identifying emerging public health issues, potentially improving the timeliness and precision of health surveillance. Insights from social media can guide public health communications and policy decisions, addressing gaps in traditional surveillance methods.

This article addresses “Precision in Prevention and Health Surveillance: How AI May Improve the Timeliness of Identification of Health Concerns through Social Media Content Analysis” by showcasing how AI-driven topic modeling can analyze large volumes of social media data to detect and understand emerging health issues. The use of transformer-based models like BERTopic allows for the precise identification of relevant topics and trends, providing real-time insights into public health concerns. This approach enhances the timeliness of health surveillance, offering early warnings and informing interventions to address new health risks, such as those associated with delta-8 THC.

No conflict of interest has been declared by the author(s).

Publication History

Article published online:
08 April 2025