Identifying clusters in Bayesian disease mapping

Abstract

Disease mapping is the field of spatial epidemiology interested in estimating the spatial pattern in disease risk across [Formula: see text] areal units. One aim is to identify units exhibiting elevated disease risks, so that public health interventions can be made. Bayesian hierarchical models with a spatially smooth conditional autoregressive prior are used for this purpose, but they cannot identify the spatial extent of high-risk clusters. Therefore, we propose a two-stage solution to this problem, with the first stage being a spatially adjusted hierarchical agglomerative clustering algorithm. This algorithm is applied to data prior to the study period, and produces [Formula: see text] potential cluster structures for the disease data. The second stage fits a separate Poisson log-linear model to the study data for each cluster structure, which allows for step-changes in risk where two clusters meet. The most appropriate cluster structure is chosen by model comparison techniques, specifically by minimizing the Deviance Information Criterion. The efficacy of the methodology is established by a simulation study, and is illustrated by a study of respiratory disease risk in Glasgow, Scotland.

Identifying Pediatric Cancer Clusters in Florida Using Log-Linear Models and Generalized Lasso Penalties

“…In the case of aerial data, frequency tests similar to those used in quadrat-based methods are frequently used (e.g., see Potthoff & Whit-tinghill, 1966a and Potthoff & Whittinghill, 1966b). Bayesian methods for disease clustering in spatially aggregated data have been proposed by Knorr-Held & Raßer (2000), Green & Richardson (2002), Wakefield & Kim (2013) and Anderson et al (2013). Other recent contributions to the field include the work of Moraga & Montes (2011), who use local indicators of spatial association (LISA) functions, Charras-Garrido et al (2012), who use a latent discrete Markov random field estimated using an expectation-maximization algorithm, and Heinzl & Tutz (2014), who propose a clustering approach that uses fused-lasso penalties to estimate the number of clusters.…”

“…On the other hand, the whole point of methods for de novo identification of cancer cluster is to pinpoint such discontinuities. Of course, these two objectives are not necessarily opposed (e.g., see Knorr-Held & Raßer, 2000, Green & Richardson, 2002 and Anderson et al, 2013), but they are certainly different.…”

A Synthesis of Spatial Models for Multivariate Count Responses

“…The model was applied to estimate the prevalence of two heart diseases across 375 boroughs in Italy's Lazio region (Alfo et al 2009), among other applications in health geography (see, e.g., Anderson et al 2014). While the finite mixture models can define clusters in a meaningful way, the models can incur excessive computation time and are considered a special type of the generalized MCAR models (Alfo et al 2009).…”

Spatio-Temporal Areal Unit Modeling in R with Conditional Autoregressive Priors Using the CARBayesST Package

“…These databases provide data on a set of K areal units for N consecutive time periods, yielding a rectangular array of K × N spatio-temporal observations. The motivations for modeling these data are varied, and include estimating the effect of a risk factor on a response (see Wakefield 2007 andLee, Ferguson, andMitchell 2009), identifying clusters of contiguous areal units that exhibit an elevated risk of disease compared with neighboring areas (see CharrasGarrido, Abrial, andde Goer 2012 andAnderson, Lee, andDean 2014), and quantifying the level of segregation in a city between two or more different groups (see Lee et al 2015). As a result different space-time structures have been proposed for modeling spatio-temporal data, which depend on the underlying question(s) of interest and the goals of the analysis.…”