Emerging poisoning trends on social media: Sodium nitrite and nitrate content on TikTok

Corazza, O, Assi, S orcid iconORCID: 0000-0002-5142-9179, Prevete, E, Venturini, S, Al-Jumeily, D and Selek, S (2026) Emerging poisoning trends on social media: Sodium nitrite and nitrate content on TikTok. Emerging Trends in Drugs, Addictions, and Health, 6. ISSN 2667-1182

[thumbnail of Emerging poisoning trends on social media_Sodium nitrite and nitrate content on TikTok.pdf]
Preview
Text
Emerging poisoning trends on social media_Sodium nitrite and nitrate content on TikTok.pdf - Published Version
Available under License Creative Commons Attribution Non-commercial No Derivatives.

Download (6MB) | Preview

Abstract

Introduction: Sodium nitrite and sodium nitrate have emerged as public health concerns due to their association with severe poisoning and intentional self-harm. As social media increasingly shapes how adolescents and young adults access and share information, platforms such as TikTok may facilitate prevention while also increasing exposure to potentially harmful narratives. This study aimed to investigate sodium nitrite/nitrate-related content on TikTok, focusing on harm-related discourse, suicide-related framing, engagement patterns, and automated harmful-content detection.

Methods: We analyzed metadata from the most-viewed TikTok videos discussing sodium nitrite using discourse analysis, BERTopic, HarmNER, BART-MNLI, DistilBERT, and supervised machine-learning models, including logistic regression and XGBoost, with SHAP-based explainability.

Results: Harm, misuse, overdose, and suicide were prominent themes. Engagement was highly skewed, with a small number of posts receiving disproportionate visibility. Prevention-oriented content was more common than explicitly harmful content, but harmful narratives remained present. Logistic regression showed the best performance for harmful-content detection.

Conclusion: TikTok contains both preventive and potentially harmful narratives related to sodium nitrite/nitrate, underscoring the platform’s dual role in self-harm communication. These findings highlight the value of explainable machine-learning approaches for the early identification of high-risk content and for strengthening moderation, suicide prevention, and digital public health responses.

Item Type: Article
Uncontrolled Keywords: 46 Information and Computing Sciences; 4701 Communication and Media Studies; 47 Language, Communication and Culture; Prevention; Machine Learning and Artificial Intelligence; Data Science; Networking and Information Technology R&D (NITRD); Suicide; Pediatric Research Initiative; Physical Injury - Accidents and Adverse Effects; Behavioral and Social Science; Suicide Prevention; Mental Health; Mental health; 3 Good Health and Well Being
Subjects: R Medicine > RS Pharmacy and materia medica
T Technology > T Technology (General)
Divisions: Computer Science and Mathematics
Pharmacy and Biomolecular Sciences
Publisher: Elsevier
Date of acceptance: 26 October 2026
Date of first compliant Open Access: 20 November 2026
Date Deposited: 20 Nov 2026 15:37
Last Modified: 20 Nov 2026 15:37
DOI or ID number: 10.1016/j.etdah.2026.100289
URI: https://researchonline.ljmu.ac.uk/id/eprint/29909
View Item View Item