NEWS ARTICLES ARE MORE LIKELY TO BE SHARED IF THEY COMBINE STATISTICS WITH EXPLANATION

Purpose: Public communications regarding vaccination have been criticized for over-reliance on statistics in the face of strident anti-vaccination narratives; thus, we conducted a retrospective observational study designed to determine how the 2014-2015 “Disneyland” measles outbreak changed the quantity, content, and persuasive effectiveness of pro- and anti-vaccination media coverage.

Method: We obtained a collection of 39,351 news articles that contained a vaccine related keyword from November 18, 2014 to March 26, 2015 using various web news search APIs. Using Amazon Mechanical Turk, we coded 2,253 articles as vaccine related/unrelated, as well as a variety of questions designed to measure perceptions of risk susceptibility and severity associated with vaccination and virus exposure. We used the Facebook API to measure the number of shares per article. We then analyzed differences in proportions of responses before and after the outbreak (Jan. 7th 2015).

Result: Following the Disneyland outbreak, more articles criticiezd those who oppose vaccination (p<0.001), independent of a total increase (p=0.002) in the total number of motivated articles devoted to vaccination. Articles were more likely to be shared on Facebook if they expressed strident opinions (e.g., editorials) compared to those that presented only objective facts (p<0.001), and articles that presented statistical information regarding vaccination rates (p=0.008) were more likely to be shared. Articles were also more likely to be shared if they emphasized the harm associated with vaccination (p<0.001), but not if they emphasized the harm associated with viral infection. Pro-vaccination articles were more likely to present statistical evidence regarding viral infection rates when compared to articles that opposed them (p=0.001). No significant differences in the use of stories or anecdotes, or in the use of statistics regarding vaccination rates, were detected between groups. Although articles supporting vaccine refusal were more likely to emphasize benefits from vaccination (p<0.001) when compared to articles that explicitly opposed pro-vaccination groups, they also emphasized possible benefits of infection by the virus (p<0.001) and harms associated with the vaccine (p=0.003). 

Conclusion: Articles were more likely to be shared if they combined statistical information with an interpretation or explanation of these data. Whereas pro-vaccination articles were unlikely to mention harms associated with vaccination or benefits from viral infection, those who oppose vaccination used a two-sided strategy that tended to acknowledge that vaccines might confer immunity.