Spatiotemporal high-resolution prediction and mapping: methodology and application to dengue disease

Area of Science:

• Epidemiology
• Geospatial analysis
• Statistical modeling

Background:

• Dengue disease poses a significant global public health challenge.
• Effective early warning systems require accurate identification and mapping of high-risk areas.
• Existing methods may not capture fine-scale spatiotemporal disease dynamics.

Purpose of the Study:

• To present a novel fusion area-cell spatiotemporal generalized geoadditive-Gaussian Markov random field (FGG-GMRF) framework.
• To enable joint estimation of complex spatiotemporal dengue risk models.
• To improve dengue risk prediction at a fine spatial resolution.

Main Methods:

• Developed the FGG-GMRF framework integrating area and cell-level data.
• Utilized spatiotemporal Gaussian fields and Gaussian Markov random fields (via FEM and SPDE) to model relative risk.
• Employed Bayesian Integrated Nested Laplace Approximation for model estimation.
• Combined low-resolution district data with high-resolution weather data.

Main Results:

• The FGG-GMRF model successfully predicted dengue relative risk at the subdistrict level in Bandung, Indonesia.
• Identified significant fine-scale heterogeneities in dengue risk not visible at the area level.
• Observed considerable temporal variations in relative risk, with seasonal patterns.

Conclusions:

• The FGG-GMRF framework provides accurate and precise dengue risk prediction at a fine spatial scale.
• This approach enhances early warning systems by revealing localized risk variations.
• The model effectively integrates multi-resolution data for improved public health surveillance.