In the presence of health threats, precision public health approaches aim to provide targeted, timely, and population-specific interventions. Accurate surveillance methodologies that can estimate infectious disease activity ahead of official healthcare-based reports, at relevant spatial resolutions, are important for achieving this goal. Here we introduce a methodological framework which dynamically combines two distinct influenza tracking techniques, using an ensemble machine learning approach, to achieve improved state-level influenza activity estimates in the United States. The two predictive techniques behind the ensemble utilize (1) a self-correcting statistical method combining influenza-related Google search frequencies, information from electronic health records, and historical flu trends within each state, and (2) a network-based approach leveraging spatio-temporal synchronicities observed in historical influenza activity across states. The ensemble considerably outperforms each component method in addition to previously proposed state-specific methods for influenza tracking, with higher correlations and lower prediction errors.

The dengue virus affects millions of people every year worldwide, causing large epidemic outbreaks that disrupt people's lives and severely strain healthcare systems. In the absence of a reliable vaccine against it or an effective treatment to manage the illness in humans, most efforts to combat dengue infections have focused on preventing its vectors, mainly the Aedes aegypti mosquito, from flourishing across the world. These mosquito-control strategies need reliable disease activity surveillance systems to be deployed. Despite significant efforts to estimate dengue incidence using a variety of data sources and methods, little work has been done to understand the relative contribution of the different data sources to improved prediction. Additionally, most work has focused on prediction systems at the national level, rather than at finer spatial resolutions. We develop a methodological framework to assess and compare dengue incidence estimates at the city level and evaluate the performance of a collection of models on 20 different cities in Brazil. The data sources we use towards this end are weekly incidence counts from prior years (seasonal autoregressive terms), weekly-aggregated weather variables, and real-time internet search data. We find that a random forest-based model effectively leverages these multiple data sources and provides robust predictions, while retaining interpretability. For real-time predictions that assume long delays (6-8 weeks) in the availability of epidemiological data, we find that real-time internet search data are the strongest predictors of Dengue incidence, whereas for predictions that assume very short delays (1-2 weeks), short-term and seasonal autocorrelation are dominant as predictors. Despite the difficulties inherent to city-level prediction, our framework achieves meaningful and actionable estimates across cities with different characteristics.