Characterizing Information Needs and Peer Engagement Regarding Medication for Opioid Use Disorder on Social Media

NIDA, 1R21DA059665-01A1

9/30/2024 – 9/29/2026

Sarah Masud Preum, PhD (Dartmouth College)

Co-I: Dr. Sarah E. Lord and Dr. Jacob T. Borodovsky (CTBH, Geisel).
PhD Student: Madhusudan Basak and Omar Sharif (Department of CS, Dartmouth).
Consultant: Dr. Edward V. Nunes (Columbia University Irving Medical Center, New York State Psychiatric Institute), Dr. Sandra Ann Springer (Yale School of Medicine).

The opioid crisis has devastating impacts, resulting in $35 billion in healthcare costs, $92 billion in lost productivity, and over 100,000 deaths annually. Effective treatment with Medications for Opioid Use Disorder (MOUD) like methadone or buprenorphine can significantly reduce opioid-related mortality. However, individuals with Opioid Use Disorder (OUD) face critical unmet information needs about MOUD treatment, often turning to social media due to stigma, lack of trust, and resources. This project aims to leverage Natural Language Processing (NLP) and mixed methodologies to identify and classify medication treatment information needs of individuals with OUD on Reddit, a popular social media platform. The goal is to extract clinically relevant insights to improve MOUD treatment access and quality.

Three specific aims guide this project:
AIM 1: Identify buprenorphine- and methadone-related TINs self-reported by persons with expressed opioid use disorder on Reddit. A large dataset of relevant Reddit posts will be curated, and a qualitative coding protocol developed to systematically identify MOUD treatment information needs.

AIM 2: Evaluate feasibility of reliably classifying buprenorphine and methadone-related TINS on Reddit using state-of-the-art NLP methods to enable efficient extraction of MOUD TINs on social media. Success will be defined by achieving an F1 score of 80% or higher and statistically significant improvements over standard baseline models.

AIM 3: Characterize the nature of peer engagement on Reddit to identify areas of MOUD misinformation and information gaps, promoted self-treatment strategies, and stigma. Both quantitative and qualitative methods will be employed to achieve this aim. The impact of this project is significant, as it will produce new, validated methods for efficiently extracting actionable insights from social media data. The outcomes align with priorities outlined by the National Institute on Drug Abuse (NIDA), including leveraging data science to understand real-world complexity and developing personalized interventions informed by people with lived experience. This work will establish a foundation for future proposals aimed at advancing computational health in the context of substance use disorders.

This project will be the first to harness advanced natural language processing to analyze social media data, revealing the treatment information needs of individuals with opioid use disorder (OUD) both within and outside traditional care settings. By identifying critical gaps, misinformation, and prevalent rumors surrounding methadone and buprenorphine treatment, this study is projected to surface valuable population-level insights related to treatment induction and adherence. These insights will inform tailored interventions and communication strategies, ultimately enhancing treatment access, adherence, and quality for those affected by OUD in a data-driven, human-centric manner.