Improved state-level influenza activity nowcasting in the United States leveraging Internet-based data sources and network approaches via ARGONet

• DOI: 10.1101/344580
• Publication Date: 2018-06-14T22:45:10Z
• Authors: Fred S. Lu; Mohammad W. Hattab; Leonardo Clemente; Mauricio Santillana
• Keywords: Nowcasting; Tracking (education)

Abstract In the presence of population-level health threats, precision public approaches seek to provide right intervention population at time. Accurate real-time surveillance methodologies that can estimate infectious disease activity ahead official healthcare-based reports, in relevant spatial resolutions, are critical eventually achieve this goal. We introduce a novel methodological framework for task which dynamically combines two distinct flu tracking techniques, using ensemble machine learning approaches, improved estimates state level US. The predictive techniques behind proposed methodology, named ARGONet, utilize (1) dynamic and self-correcting statistical approach combine flu-related Google search frequencies, information from electronic records, historical trends within given state, as well (2) data-driven network-based leverages temporal synchronicities observed across states improve state-level estimates. considerably outperforms each individual method any previously state-specific tracking, with higher correlations lower prediction errors.