Epidemic State Estimation with Syndromic Surveillance and ILI Data Using Particle Filter

Designing effective mitigation strategies against an influenza outbreak requires an accurate prediction of a disease’s future course of spreading. Real time information such as syndromic surveillance data and reporting of influenza cases by clinicians can be used to generate an estimate of the current state of spreading of a disease. Syndromic surveillance data is immediately available compared to clinical reports that require data collection and processing. On the other hand, syndromic data is less credible than the clinically confirmed case reports. In this paper, we present a method to combine immediately-available-but-highly-uncertain syndromic surveillance data with credible-but-time-delayed clinical case report data. This problem is formulated as a non-linear stochastic filtering problem and solved by a particle filtering method. Our experimental results on a hypothetical pandemic scenario show that state estimation is improved by utilizing both data sets than when using only one of them, but the amount of improvement depends on relative credibility and length of delay of clinical case report data. This result is explained with a preliminary analysis for a linear, Gaussian case.