Abstract
This study examines the patterns of medical preprint sharing on Twitter during the early stages of the COVID-19 pandemic. Our analysis demonstrates a stark increase in attention to medical preprints among the general public since the beginning of the pandemic. We also observe a political divide in medical preprint sharing patterns—a finding in line with previous observations regarding the politicization of COVID-19-related discussions. In addition, we find that the increase in attention to preprints from members of the general public coincided with a significant change in the social media-based discourse surrounding these documents.
Introduction
Traditionally, medical preprints—manuscripts shared publicly before undergoing formal peer review—were the exclusive domain of specialists. These researchers understood the "handle with care" nature of unvetted data. However, the urgency of the COVID-19 pandemic fundamentally altered this ecosystem. Platforms like medRxiv became essential for the rapid dissemination of life-saving data, but this acceleration brought unintended consequences: the weaponization of preliminary findings to support political narratives.
This paper investigates the "politicization" of medical information by quantifying the shift in audience and discourse on Twitter. By analyzing how different ideological groups engaged with unvetted research, we reveal how the pandemic transformed preprints from niche scientific tools into mainstream political instruments.
Methodology
To track this shift, we employed a quantitative descriptive approach combining social media analytics and natural language processing. Our primary dataset consisted of 557,405 tweets containing medRxiv URLs posted between June 2019 and June 2020.
User Classification and Metrics
We categorized Twitter users into four primary groups—Democrat, Republican, Scientist, and Medical Professional—based on keyword matching in user biographies (e.g., "GOP" or "MAGA" for Republicans; "PhD" or "University" for scientists). To ensure accuracy, we cross-validated this classification using ideology detection methods based on following patterns and checked for automated manipulation using Botometer.
Two key metrics were developed to analyze the data:
- Normalized Share of Interest: This compared the sharing patterns of our sample against a baseline of approximately 850,000 random Twitter users to avoid bias from the general popularity of certain groups.
- Party Share Index (PSI): A log-transformed metric designed to measure the ratio of Republican-to-Democrat sharing for specific preprints.
Key Findings
The Linguistic Shift: From "Preprint" to "Study"
One of the most significant findings is the change in how users described the documents they shared. Before the pandemic, the term "preprint" was common, signaling an awareness of the document's preliminary status. As the pandemic progressed, the term "study" became dominant among the general public. This linguistic shift suggests that unvetted manuscripts were increasingly treated as established scientific facts rather than works in progress.
Partisan Asymmetry and Topic-Driven Engagement
The sharing of preprints was not equally distributed across the political spectrum. Our sample was heavily skewed toward conservative-leaning users, who shared medRxiv links at a rate significantly higher than their presence in the random baseline sample.
This partisanship was highly topic-specific. Preprints regarding hydroxychloroquine (HCQ) and the Santa Clara antibody seroprevalence study showed the highest levels of Republican-leaning partisanship. For example, one study on HCQ and azithromycin reached a PSI of 6.247, indicating it was shared almost exclusively within Republican-leaning circles.
The Dilution of Expertise
Early in the pandemic, medical professionals and scientists were the primary drivers of preprint discourse. However, as the pandemic reached its first major wave in April 2020, their "share of voice" was diluted by a massive influx of partisan users. This surge effectively drowned out nuanced scientific discussion with political framing. Notably, bot activity was found to be marginal (approximately 3%), suggesting that this politicization was driven by real users rather than automated manipulation.
Discussion and Limitations
While our study provides a clear picture of the politicization of science on Twitter, it is limited by its focus on a single platform and the American political context. Identifying political orientation via Twitter bios is a proxy that, while validated, remains an aggregate measure. Furthermore, discourse on platforms like Facebook or Telegram may exhibit different dynamics.
Conclusion
The transition of preprints from scientific artifacts to political tools highlights a growing challenge in science communication. When preliminary research is used to bolster political arguments before it can be verified, the integrity of public health discourse is at risk. We suggest that social media platforms consider implementing specific labels for preprints—similar to those used for state-affiliated media—to help the general public distinguish between vetted and unvetted scientific findings.
Summary updated