Poisson Cokriging for Public Health Response
David Payares Garcia is a PhD student in the Department of Earth Observation Science. (Co)Promotors are prof.dr. A. Stein and dr. F.B. Osei from the Faculty ITC and prof.dr. J. Mateu from the University Jaume I.
Spatial disease mapping is critical for public health surveillance, yet traditional methods struggle with discrete count data and fail to exploit relationships between related diseases. This thesis introduces Poisson cokriging, a comprehensive geostatistical framework for multivariate disease mapping that addresses fundamental challenges in epidemiological analysis.
Poisson cokriging handles count data that violate Gaussian assumptions, stabilizes disease rates in sparsely populated areas, and leverages information from multiple diseases simultaneously. The method incorporates population-weighted estimators and bias correction to account for data heteroscedasticity while maintaining computational efficiency.
The thesis systematically develops Poisson cokriging through multiple extensions. The bivariate formulation demonstrates 50-74% improvement in prediction accuracy over univariate methods. Multivariate extensions enable simultaneous analysis of multiple diseases with superlinear improvements. Spatial downscaling through Area-to-Area and Area-to-Point approaches generates high-resolution risk maps from aggregated data with over 98% accuracy. The spatio-temporal extension captures disease dynamics across space and time.
Extensive simulation studies confirm robust statistical properties and exceptional computational efficiency, with analyses completing in seconds compared to hours required by comparable Bayesian methods. Real-world applications demonstrate the method's versatility: arboviral disease mapping in Colombia revealed vector dispersal patterns; COVID-19 analysis in the Netherlands and Colombia uncovered syndemic relationships with chronic diseases; HIV/STD mapping in Pennsylvania effectively handled sparse data; and suicide rate analysis across the United States identified continental-scale patterns informing national prevention strategies.
Poisson cokriging bridges methodological innovation and practical application, enabling accurate disease mapping in diverse settings, identifying syndemic patterns, supporting real-time surveillance, and providing uncertainty quantification for resource allocation. The method equips health agencies with rigorous yet computationally feasible tools for modern disease surveillance and intervention planning.
