Mining Social Media and Web Searches For Disease Detection

Yang, Y. Tony; Horneffer, Michael; DiLisio, Nicole

The web-based search query data accurately predicted the increase in ILI percentages. Such data can be acquired, processed, and analysed quickly. As a result, the ILI estimates found by Ginsberg et al. were consistently 1-2 weeks ahead of CDC ILI surveillance reports. 8 Search query-based predictive models may give public health professionals a greater advantage in developing earlier responses to seasonal outbreaks of flu. Due to the widespread utilisation and popularity of the Google search engine, Google-based search query analytics is one of the most timely and broad-reaching flu monitoring systems available today. Like other web-based surveillance systems, however, Ginsberg et al.'s system remains susceptible to false alerts resulting from ILI inquiries unrelated to the actual occurrence of flu in any one population group. Therefore, human analysis and interpretation of search query data remains necessary.