State Legislators’ Divergent Social Media Response to the Opioid Epidemic from 2014 to 2019: Longitudinal Topic Modeling Analysis

Abstract

Background: The opioid epidemic is widely recognized as a legislative priority, but there is substantial variation in state adoption of evidence-based policy. State legislators’ use of social media to disseminate information and to indicate support for specific initiatives continues to grow and may reflect legislators’ openness to opioid-related policy change. Objective: We sought to identify changes in the national dialogue regarding the opioid epidemic among Democratic and Republican state legislators and to estimate changing partisanship around understanding and addressing the epidemic over time. Design: Longitudinal natural language processing analysis. Participants: A total of 4083 US state legislators in office between 2014 and 2019 with any opioid-related social media posts. Main Measures: Association between opioid-related post volume and state overdose mortality, as measured by Kendall’s rank correlation coefficient. Latent Dirichlet allocation analysis of all social media posts to identify key opioid-related topics. Longitudinal analysis of differences in the prevalence of key topics among Democrats and Republicans over time. Key Results: In total, 43,558 social media posts met inclusion criteria, with the vast majority to Twitter (n=28,564; 65.6%) or Facebook (n=14,283; 32.8%). Posts were more likely to mention fentanyl and less likely to mention heroin over time. The volume of opioid-related content was positively associated with state-level unintentional overdose mortality among both Democrats (tau=0.42, P