Forecasting seasonal outbreaks of influenza

Abstract

Influenza recurs seasonally in temperate regions of the world; however, our ability to predict the timing, duration, and magnitude of local seasonal outbreaks of influenza remains limited. Here we develop a framework for initializing real-time forecasts of seasonal influenza outbreaks, using a data assimilation technique commonly applied in numerical weather prediction. The availability of realtime, web-based estimates of local influenza infection rates makes this type of quantitative forecasting possible. Retrospective ensemble forecasts are generated on a weekly basis following assimilation of these web-based estimates for the 2003-2008 influenza seasons in New York City. The findings indicate that real-time skillful predictions of peak timing can be made more than 7 wk in advance of the actual peak. In addition, confidence in those predictions can be inferred from the spread of the forecast ensemble. This work represents an initial step in the development of a statistically rigorous system for real-time forecast of seasonal influenza.Kalman filter | absolute humidity W orldwide, influenza produces 3-5 million severe illnesses annually and kills an estimated 250,000-500,000 people (1). In temperate regions, influenza characteristically recurs during winter when absolute humidity levels are low (2, 3), but at present our ability to predict important details of these seasonal influenza outbreaks is limited. Indeed, much public health benefit could be gleaned from early, skillful prediction of the onset, peak, duration, and magnitude of local influenza outbreaks.Mathematical models of infectious disease transmission have been in use for over a century (4). These models have been developed to study the dynamic properties of disease transmission (5-7), determine the biological characteristics of specific pathogens (8,9), and analyze historical transmission behavior during documented outbreak events (10).More recently, infectious disease model simulations have been performed retrospectively in conjunction with statistical filtering methods to provide maximum-likelihood parameter estimation (11, 12) and improved epidemic simulation through time and physical space (13-16). Filtering techniques iteratively update, or adjust, model simulation estimates of the dynamic state, e.g., population infection rates, using real-world observations of that state, as the model is integrated through time. Because the state is only intermittently, or partially, observed-i.e., infections may be observed only for some locations and times, and some state variables, such as population susceptibility rates, may not be observed at all-and because these partial observations themselves contain error, the filter endeavors to balance the relative information contained in the observations and the model simulation. At the same time, the filtering process can also be used to estimate epidemiologically significant parameters within a model. These same filtering techniques, by constraining the model state and parameters, can potentially be used to e...

Forecasting influenza outbreak dynamics in Melbourne from Internet search query surveillance data

“…Shaman et al 4, 5. used an ensemble adjustment Kalman filter (EAKF) and an SIRS infection model to predict seasonal influenza outbreaks in 108 cities across the USA and showed that outbreak peak timing could be predicted 4–6 weeks in advance.…”

“…A variety of recursive Bayesian estimation methods (‘filters’) have been used for such forecasting purposes,1, 2 often in combination with Internet search query surveillance data and mechanistic models of infection 3, 4, 5, 6, 7, 8…”

Digital epidemiology reveals global childhood disease seasonality and the effects of immunization

“…Internet query platforms, such as Google Trends, have provided powerful and accessible resources for identifying outbreaks and for implementing intervention strategies (12)(13)(14). Research on infectious disease informationseeking behavior has demonstrated that Internet queries can complement traditional surveillance by providing a rapid and efficient means of obtaining large epidemiological datasets (13,(15)(16)(17)(18). For example, epidemiological information contained within Google Trends has been used in the study of rotavirus, norovirus, and influenza (14,15,17,18).…”

Accurate estimation of influenza epidemics using Google search data via ARGO

“…Our ability to effectively prepare for and respond to these outbreaks heavily relies on the availability of accurate realtime estimates of their activity. Existing methods to predict the timing, duration, and magnitude of flu outbreaks remain limited (18). Well-established clinical methods to track flu activity, such as the CDC's ILINet, report the percentage of patients seeking medical attention with ILI symptoms (www.cdc.gov/flu/).…”

“…This time lag is far from optimal for decision-making purposes. To alleviate this information gap, multiple methods combining climate, demographic, and epidemiological data with mathematical models have been proposed for real-time estimation of flu activity (18,(21)(22)(23)(24)(25). In recent years, methods that harness Internet-based information have also been proposed, such as Google (1), Yahoo (2), and Baidu (3) Internet searches, Twitter posts (4), Wikipedia article views (5), clinicians' queries (6), and crowdsourced selfreporting mobile apps such as Influenzanet (Europe) (26), Flutracking (Australia) (27), and Flu Near You (United States) (28).…”