Speaker
Description
Following the COVID-19 pandemic, seasonal infectious disease trend shifts have highlighted the need for advanced early warning systems. Conventional methods, such as change point detection and hockey-stick regression, are widely used but are designed for retrospective analysis. Building on these approaches, we aim to develop a generalized real-time early detection model based on bootstrap clustering for various seasonal infectious diseases.
This study utilizes influenza-like illness (ILI), respiratory syncytial virus (RSV), norovirus, and hand-foot-and-mouth disease (HFMD) from the Korea Disease Control and Prevention Agency (KDCA). Bootstrap clustering traditionally evaluates stability via the Jaccard coefficient \cite{Yu2019}. In this study, we extend this approach beyond stability assessment to develop a distribution-based early detection framework. Furthermore, we conducted real-time detection simulations by sequentially incorporating weekly data. This approach supports a three-level warning system consisting of Caution, Alert, and Severe stages using distribution-based early detection.
This framework provides a generalized early detection model that can be applied to various seasonal infectious diseases. Unlike conventional statistical methods, the proposed approach enables real-time early detection. By introducing a three-level warning system, this approach has the potential to enhance timely preparedness and response for a wide range of seasonal infectious diseases.
Bibliography
@article{Yu2019, title={Bootstrapping estimates of stability for clusters, observations and model selection}, volume={34}, ISSN={0943-4062, 1613-9658}, url={http://link.springer.com/10.1007/s00180-018-0830-y}, DOI={10.1007/s00180-018-0830-y}, number={1}, journal={Computational Statistics}, author={Yu, Han and Chapman, Brian and Di Florio, Arianna and Eischen, Ellen and Gotz, David and Jacob, Mathews and Blair, Rachael Hageman}, year={2019}, month=mar, pages={349--372}, language={en} }
