Using networks to combine 'big data' and traditional surveillance to improve influenza predictions

Abstract

Seasonal influenza infects roughly 5–20 percent of the U.S. population each year, causing more than 200,000 hospitalizations. Accurately assessing infection levels and predicting regional risk can guide targeted prevention and treatment, especially during epidemics. Google Flu Trends (GFT) sparked optimism that big data could estimate disease burden in real time, producing measures two weeks earlier than CDC surveillance. However, GFT suffered notable errors and is substantially less accurate at tracking laboratory-confirmed cases than syndromic influenza-like illness (ILI) measures. We construct an empirical network using CDC data and combine it with GFT to markedly improve predictive performance. The combined model predicts infections one week into the future as well as GFT predicts the present and performs particularly well in highly connected regions and during epidemic periods.

Cite

@article{davidson2015using,
title={Using networks to combine "big data" and traditional surveillance to improve influenza predictions},
author={Davidson, Michael W and Haim, Dotan A and Radin, Jennifer M},
journal={Scientific Reports},
volume={5},
number={1},
pages={8154},
year={2015},
publisher={Nature Publishing Group UK}
}