State-space modelling for infectious disease surveillance data: Dynamic regression and covariance analysis

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• Understanding infectious disease transmission dynamics is crucial for effective public health strategies.
• Traditional surveillance methods may not fully capture the complex factors influencing disease spread.
• Ontario, Canada, provides a rich dataset for analyzing COVID-19 transmission patterns.

Purpose of the Study:

• To apply advanced state-space modeling techniques to analyze COVID-19 surveillance data from Ontario.
• To investigate the relationships between COVID-19 cases, hospitalizations, workdays, and wastewater viral loads.
• To develop novel methods for time-varying correlation estimation in non-stationary epidemiological data.

Main Methods:

• Component linear Gaussian state-space models were used to analyze case counts, capturing periodicity, trends, and fluctuations.
• Dynamic regression models explored the influence of covariates like workdays and wastewater viral loads.
• A novel methodology was employed for multivariate covariance estimation, providing time-varying correlation estimates.

Main Results:

• State-space models demonstrated superior performance in capturing disease transmission dynamics compared to simpler models.
• Environmental covariates, particularly wastewater data, were significant in enhancing model robustness.
• The study uncovered complex interplays between epidemiological factors and public health outcomes.

Conclusions:

• Advanced state-space modeling offers significant advantages for analyzing dynamic infectious disease data.
• Integrating environmental surveillance data, such as wastewater, is vital for improving epidemiological models.
• This research provides a deeper understanding of disease transmission, informing targeted public health interventions.