Forecasting Seasonal Influenza Fusing Digital Indicators and a Mechanistic Disease Model

Q, Zhang, N. Perra, D. Perrotta, D. Paolotti, M. Tizzoni, A. Vespignani.
WWW '17
pp 311-319
April 3, 2017


The availability of  novel digital data streams that can be used as proxy for monitoring  infectious disease incidence is ushering in a new era for real-time forecast  approaches to disease spreading. Here, we propose the first seasonal  influenza forecast framework based on a stochastic, spatially structured  mechanistic model (individual level microsimulation) initialized with  geo-localized microblogging data. The framework provides for more than 600  census areas in the United States, Italy and Spain, the initial conditions  for a stochastic epidemic computational model that generates an ensemble of  forecasts for the main indicators of the epidemic season: peak time and  intensity. We evaluate the forecasts accuracy and reliability by comparing  the results with the data from the official influenza surveillance systems in  the US, Italy and Spain in the seasons 2014/15 and 2015/16. In all countries  studied, the proposed framework provides reliable results with leads of up to  6 weeks that became more stable and accurate with progression of the season.  The results for the United States have been generated in real-time in the  context of the Centers for Disease Control and Prevention ``Forecasting the  Influenza Season Challenge''. A characteristic feature of the mechanistic  modeling approach is in the explicit estimate of key epidemiological  parameters relevant for public health decision-making that cannot be achieved  with statistical models that do not consider the disease dynamic.  Furthermore, the presented framework allows the fusion of multiple data  streams in the initialization stage and can be enriched with census, weather  and socioeconomic data.