Spatial transmission network construction of influenza-like illness using dynamic Bayesian network and...

Area of Science:

• Epidemiology
• Network Science
• Public Health

Background:

• Influenza epidemics necessitate informed prevention decisions for optimal resource allocation.
• Understanding spatio-temporal transmission patterns is crucial for effective influenza control.
• Characterizing influenza spread among adjacent regions aids in predicting epidemic dynamics.

Purpose of the Study:

• To propose a novel concept of spatio-temporal routes for constructing influenza transmission networks.
• To develop methods for estimating spatio-temporal routes and transmission impacts.
• To visualize and quantify influenza transmission pathways between geographical locations.

Main Methods:

• Utilized influenza-like illness (ILI) data from 21 cities in Sichuan province (2010-2016).
• Employed a joint dynamic Bayesian network (DBN) and vector autoregressive moving average (VARMA) model for route estimation.
• Applied structure learning (first-order conditional dependencies) and parameter learning (VARMA) to build the transmission network.

Main Results:

• Identified longer influenza transmission cycles in Western Sichuan and Chengdu Plain compared to Northeastern Sichuan.
• Revealed potential spatio-temporal routes of influenza entry into Sichuan from bordering regions.
• Highlighted specific spatio-temporal routes with strong associations, indicating potential surveillance hot spots.

Conclusions:

• Introduced a new framework for identifying stable spatio-temporal disease transmission routes.
• Enabled measurement of transmission effect sizes between locations.
• Aids in timely prediction of infectious disease trends and identification of key epidemic areas.