Source-linked AI summary
Tweeting biomedicine: an analysis of tweets and citations in the biomedical literature
Stefanie Haustein, Isabella Peters, Cassidy R. Sugimoto, Mike Thelwall, Vincent Larivière
TL;DR
This paper examines Twitter’s use for disseminating biomedical papers and its relationship to traditional citation indicators. Using large-scale biomedical publication, tweet, and citation data, it finds low overall coverage, substantial disciplinary variation, and low tweet–citation correlations, motivating a framework for interpreting Twitter-based metrics.
Problem
The paper examines how widely biomedical papers appear on Twitter, how this varies across journals and fields, and whether tweets relate to citations.
Method
The study analyzes Twitter mentions and Web of Science citations for more than 1.4 million PubMed and WoS documents, comparing coverage and relationships across journals, disciplines, and specialties.
Results
Less than 10% of PubMed documents were mentioned on Twitter, tweeting varied substantially across disciplines and specialties, and the document-level tweet–citation correlation was .183.
Takeaways & Limitations
Twitter-based indicators should complement rather than replace citation-based indicators because they capture a different type of impact, and field differences require accounting when comparing articles.
Takeaways & Limitations
Twitter citations do not reflect traditional research impact, potentially because Twitter uptake among scientists is low and scholarly use of the platform remains uncertain.
Abstract
from arXiv · showhide
Data collected by social media platforms have recently been introduced as a new source for indicators to help measure the impact of scholarly research in ways that are complementary to traditional citation-based indicators. Data generated from social media activities related to scholarly content can be used to reflect broad types of impact. This paper aims to provide systematic evidence regarding how often Twitter is used to diffuse journal articles in the biomedical and life sciences. The analysis is based on a set of 1.4 million documents covered by both PubMed and Web of Science (WoS) and published between 2010 and 2012. The number of tweets containing links to these documents was analyzed to evaluate the degree to which certain journals, disciplines, and specialties were represented on Twitter. It is shown that, with less than 10% of PubMed articles mentioned on Twitter, its uptake is low in general. The relationship between tweets and WoS citations was examined for each document at the level of journals and specialties. The results show that tweeting behavior varies between journals and specialties and correlations between tweets and citations are low, implying that impact metrics based on tweets are different from those based on citations. A framework utilizing the coverage of articles and the correlation between Twitter mentions and citations is proposed to facilitate the evaluation of novel social-media based metrics and to shed light on the question in how far the number of tweets is a valid metric to measure research impact.
Discussion and Framework
Twitter coverage of biomedical literature was generally low and varied substantially across disciplines, specialties, and journals. Low tweet–citation correlations support treating Twitter indicators as complementary rather than alternative measures, while the framework highlights multiple interpretations and unresolved questions about tweeting motivations.
- Twitter coverage: Less than 10% of more than 1.4 million PubMed–WoS articles were tweeted, although over 20% of 2012 articles received at least one tweet.Most journals had less than 20% of their content tweeted; tweeted articles averaged 2.5 tweets, while the overall rate including untweeted articles was 0.2 tweets per article.
- Variation across fields: Twitter coverage and citation rates varied widely across disciplines and specialties, requiring field differences to be considered when comparing social-media impact.Some journals and specialties attracted greater Twitter interest, while official journal or publisher handles did not necessarily increase citation rates.
- Tweets and citations: Twitter–citation relationships were positive but low: journal-level correlations with bibliometric indicators ranged from .223 to .312, and document-level correlation was .183.These findings indicate that tweets and citations are related but mostly measure different types of impact.
- Framework: The paper proposes a four-case framework classifying tweeting and citing relationships by Twitter coverage and tweet–citation correlation.The cases distinguish combinations such as high coverage with positive correlation, suggesting similar interests between Twitter users and the scientific community, and high coverage with negative correlation, suggesting divergent interests.
- Interpretation: The framework treats Twitter indicators as complementary to citation indicators because they reflect different, multifaceted aspects of research impact.The paper argues that altmetrics such as Twitter dissemination, Mendeley readership, and F1000 peer review should be analyzed separately rather than aggregated into one number.
- Limitations and future research: Further research must identify who tweets biomedical papers and why, because attention may reflect health relevance, topicality, humor, curiosity, or journal curation rather than intellectual impact.The authors call for qualitative Twitter-content analyses, user surveys, and studies beyond medical disciplines before tweets are incorporated into research evaluation.