Quantifying the effect of media limitations on outbreak data in a global online web-crawling epidemic intelligence system, 2008-2011.

Abstract

This is the first study quantitatively evaluating the effect that media-related limitations have on data from an automated epidemic intelligence system. We modeled time series of HealthMap's two main data feeds, Google News and Moreover, to test for evidence of two potential limitations: first, human resources constraints, and second, high-profile outbreaks "crowding out" coverage of other infectious diseases. Google News events declined by 58.3%, 65.9%, and 14.7% on Saturday, Sunday and Monday, respectively, relative to other weekdays. Events were reduced by 27.4% during Christmas/New Years weeks and 33.6% lower during American Thanksgiving week than during an average week for Google News. Moreover data yielded similar results with the addition of Memorial Day (US) being associated with a 36.2% reduction in events. Other holiday effects were not statistically significant. We found evidence for a crowd out phenomenon for influenza/H1N1, where a 50% increase in influenza events corresponded with a 4% decline in other disease events for Google News only. Other prominent diseases in this database - avian influenza (H5N1), cholera, or foodborne illness - were not associated with a crowd out phenomenon. These results provide quantitative evidence for the limited impact of editorial biases on HealthMap's web-crawling epidemic intelligence.