Time series modeling of pathogen-specific disease probabilities with subsampled data.

Area of Science:

• Epidemiology
• Biostatistics
• Infectious Disease Modeling

Background:

• Diseases often result from exposure to multiple pathogens, with varying clinical severity.
• Modeling pathogen-specific disease incidence over time is crucial for public health surveillance.
• Traditional Bayesian methods for pathogen imputation are computationally intensive.

Purpose of the Study:

• To develop a practical and computationally feasible Bayesian approach for modeling disease incidence probabilities given pathogen type.
• To investigate the association between time-varying meteorological covariates and disease incidence for specific pathogens.
• To analyze hand, foot, and mouth disease (HFMD) in China, focusing on enterovirus 71 (EV71) and Coxackie A16 (CA16).

Main Methods:

• Utilized an empirical Bayes procedure for initial estimation of summary statistics.
• Developed a Bayesian generalized additive model using estimated summary statistics as observed data.
• Employed a penalized B-spline model with random effects to account for time confounding and smooth covariate effects.

Main Results:

• Identified significant associations between EV71 and CA16 pathogens and meteorological factors including temperature, relative humidity, and wind speed.
• Observed similar functional relationships between these weather variables and both EV71 and CA16.
• Successfully modeled time-varying confounding using penalized B-splines, with smoothing determined by prior selection.

Conclusions:

• The proposed empirical Bayes and Bayesian generalized additive model approach offers a practical alternative to computationally challenging imputation methods.
• Meteorological factors play a significant role in the incidence of EV71 and CA16, key pathogens for HFMD.
• The model effectively handles complex epidemiological data, providing insights into pathogen-disease relationships and environmental influences.