Accurate Influenza Monitoring and Forecasting Using Novel Internet Data Streams: A Case Study in the Boston Metropolis.

摘要: Background: Influenza outbreaks pose major challenges to public health around the world, leading thousands of deaths a year in United States alone. Accurate systems that track influenza activity at city level are necessary provide actionable information can be used for clinical, hospital, and community outbreak preparation. Objective: Although Internet-based real-time data sources such as Google searches tweets have been successfully produce estimates ahead traditional care–based national state levels, tracking forecasting finer spatial resolutions, level, remain an open question. Our study aimed present precise, near methodology capable producing those collected published by Boston Public Health Commission (BPHC) metropolitan area. This approach has great potential extended other cities with access similar sources. Methods: We first tested ability searches, Twitter posts, electronic records, crowd-sourced reporting system detect metropolis separately. then adapted multivariate dynamic regression method named ARGO (autoregression general online information), designed showed it effectively uses above monitor forecast 1 week current date. Finally, we presented ensemble-based combining from models based on multiple more robustly nowcast well The performances our were evaluated out-of-sample fashion over 4 seasons within 2012-2016, holdout validation period 2016 2017. Results: methods incorporating diverse sources, including ARGO, produced most robust accurate results. observed Pearson correlations between flu historically reported BPHC 0.98 nowcasting 0.94 Conclusions: show when combined using informed, methodology, early indicators fine geographic resolutions.