Modeling seasonality in space-time infectious disease surveillance data

Area of Science:

• Epidemiology
• Biostatistics
• Time Series Analysis

Background:

• Infectious disease surveillance data often presents as multivariate time series within geographical regions.
• Seasonal variations in disease notification are common and require appropriate modeling for accurate analysis.
• Understanding disease transmission dynamics necessitates robust spatiotemporal modeling.

Purpose of the Study:

• To extend existing time series models for spatiotemporal surveillance counts.
• To incorporate seasonal variation across three distinct components within the model.
• To evaluate the model's performance in identifying seasonality and improving predictions.

Main Methods:

• Development of an extended time series model for spatiotemporal disease counts.
• Inclusion of three distinct components to capture different types of seasonality.
• Conducting a simulation study to assess model identifiability and predictive performance.
• Application of the model to real-world influenza surveillance data from Southern Germany.

Main Results:

• Simulation results demonstrate the identifiability of different seasonality types.
• The proposed model selection approach shows good predictive performance.
• Applying the model to influenza data resulted in a better model fit.
• One-step-ahead predictions for influenza spread were significantly improved.

Conclusions:

• Incorporating seasonal variation into spatiotemporal models is crucial for accurate infectious disease surveillance.
• The extended model effectively captures and accounts for multiple facets of seasonality.
• This approach enhances understanding and prediction of disease transmission patterns.
• Improved model fit and predictive accuracy were achieved for influenza surveillance data.