Spatial Downscaling of Multivariate Disease Risk

Downscaling areal health data to a finer resolution is important for understanding the intricate spatial patterns of disease. It helps to identify shared risk factors and to develop targeted public health interventions. This paper introduces Area-to-Area (ATA) and Area-to-Point (ATP) Poisson cokriging for downscaling spatial disease risks from aggregated areal data. The methodology addresses key challenges by incorporating correlation between the diseases, adjusting for population heterogeneity, and the varying shapes and sizes of the spatial entities. Simulation studies demonstrate the superior performance of ATA and ATP Poisson cokriging compared to their univariate counterparts. We achieved lower mean squared prediction errors and better preserved small-scale spatial variations. The methods are applied to COVID-19 and asthma occurrences in Bogota, Colombia. They reveal more detailed hotspots and coldspots and refined estimates of COVID-19 risk by leveraging its correlation with asthma. Our methods offer advantages in multivariate disease mapping by enabling more accurate risk assessment, improved small-area estimation, and enhanced understanding of spatial disease patterns. Their ability to downscale risks for multiple diseases simultaneously provides valuable insights for targeted public health interventions and resource allocation.