Disease mapping models for data with weak spatial dependence or spatial discontinuities

Helena Baptista,Peter Congdon,Sara S Dias,Helena Canhão,Jorge M Mendes,Ana M Rodrigues

doi:10.1515/em-2019-0025

Abstract

Abstract Recent advances in the spatial epidemiology literature have extended traditional approaches by including determinant disease factors that allow for non-local smoothing and/or non-spatial smoothing. In this article, two of those approaches are compared and are further extended to areas of high interest from the public health perspective. These are a conditionally specified Gaussian random field model, using a similarity-based non-spatial weight matrix to facilitate non-spatial smoothing in Bayesian disease mapping; and a spatially adaptive conditional autoregressive prior model. The methods are specially design to handle cases when there is no evidence of positive spatial correlation or the appropriate mix between local and global smoothing is not constant across the region being study. Both approaches proposed in this article are producing results consistent with the published knowledge, and are increasing the accuracy to clearly determine areas of high- or low-risk.

Highlights

To allocate the scarce health resources to the spatial units that need them the most is of paramount importance nowadays
Methods to identify excess risk in particular areas should ideally acknowledge and examine the extent of potential spatial clustering in health outcomes (Tosetti et al 2018).Identification of risk may be based on relatively rare area health outcomes, and model based methods are required for spatial smoothing, typically using Bayesian principles (Best, Richardson, and Thomson 2005).Where it exists, spatial clustering is the basis for local smoothing, or spatial borrowing of strength
The inverse happens in two other areas, “Tâmega” and “Alto Trás-os-Montes”. Both areas have a low value of per capita purchasing power index (PcPp), which would indicate a high risk for the disease, but LLB model posterior standardised morbidity ratio (SMR) consider those as low-risk areas

Summary

Introduction

To allocate the scarce health resources to the spatial units that need them the most is of paramount importance nowadays. In cases of epidemiological studies with relatively small sample sizes in some (almost all) of the areas, the classical estimators of the morbidity rates show high variability, and spatial disease mapping models overcome that by borrowing strength from spatial neighbours. The rationale of our approach is the following: in cases of diseases with no environmental determinant factors, use of a positive spatial correlation based on physical distance or adjacency, in the GRF/GMRF model, may not be the best way to reflect similarity between areas.

Results

Conclusion