Bayesian disease mapping: Past, present, and future

MacNab, Ying C.

doi:10.1016/j.spasta.2022.100593

Cited by 20 publications

(43 citation statements)

References 129 publications

(247 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We name hereafter B the spatial dependence matrix and the coefficients falsefalse{ B i k , normal∀ k ∼ i falsefalse} of false0 E false( ψ i falsefalse| bold-italicψ − i false) in (5) the coefficients of influence . 18…”

Section: Conditionally Formulated Gaussian Markov Random Fieldsmentioning

confidence: 99%

“…We name hereafter B the spatial dependence matrix and the coefficients {B ik , ∀k ∼ i} of E(ψ i |ψ −i ) in (5) the coefficients of influence. 18 The GMRF precision matrix (6) must be symmetric and non-negative definite. To fulfil the two requirements, functional characterizations and simplified parameterizations to the CAR conditionals have been proposed (mainly) in the disease mapping literature; and the previously mentioned iCAR, pCAR, and LCAR are the most commonly used GMRFs in disease mapping applications at the present time; see Table 1 for the CAR specifications, where key references are given.…”

Section: A Car Construction and Its Three Options Of Parameterizationmentioning

confidence: 99%

See 1 more Smart Citation

Revisiting Gaussian Markov random fields and Bayesian disease mapping

MacNab

2022

Stat Methods Med Res

Self Cite

View full text Add to dashboard Cite

We revisit several conditionally formulated Gaussian Markov random fields, known as the intrinsic conditional autoregressive model, the proper conditional autoregressive model, and the Leroux et al. conditional autoregressive model, as well as convolution models such as the well known Besag, York and Mollie model, its (adaptive) re-parameterization, and its scaled alternatives, for their roles of modelling underlying spatial risks in Bayesian disease mapping. Analytic and simulation studies, with graphic visualizations, and disease mapping case studies, present insights and critique on these models for their nature and capacities in characterizing spatial dependencies, local influences, and spatial covariance and correlation functions, and in facilitating stabilized and efficient posterior risk prediction and inference. It is illustrated that these models are Gaussian (Markov) random fields of different spatial dependence, local influence, and (covariance) correlation functions and can play different and complementary roles in Bayesian disease mapping applications.

show abstract

Section: Conditionally Formulated Gaussian Markov Random Fieldsmentioning

confidence: 99%

Section: A Car Construction and Its Three Options Of Parameterizationmentioning

confidence: 99%

Revisiting Gaussian Markov random fields and Bayesian disease mapping

MacNab

2022

Stat Methods Med Res

Self Cite

View full text Add to dashboard Cite

show abstract

“…Disease mapping is an important statistical tool used in epidemiology to explore spatial variation in disease incidence rates. Disease mapping models can generate and test hypotheses about associations between disease and a variety of potential explanatory variables, such as environmental and socio-economic factors [2,13]. Typically, disease counts, y i ( i = 1, …, n ), are collected across a study area separated into n contiguous areas.…”

Section: Modelling Approachmentioning

confidence: 99%

A Bayesian modelling framework to quantify multiple sources of spatial variation for disease mapping

Lee

Economou

Lowe

2022

J. R. Soc. Interface.

View full text Add to dashboard Cite

Spatial connectivity is an important consideration when modelling infectious disease data across a geographical region. Connectivity can arise for many reasons, including shared characteristics between regions and human or vector movement. Bayesian hierarchical models include structured random effects to account for spatial connectivity. However, conventional approaches require the spatial structure to be fully defined prior to model fitting. By applying penalized smoothing splines to coordinates, we create two-dimensional smooth surfaces describing the spatial structure of the data while making minimal assumptions about the structure. The result is a non-stationary surface which is setting specific. These surfaces can be incorporated into a hierarchical modelling framework and interpreted similarly to traditional random effects. Through simulation studies, we show that the splines can be applied to any symmetric continuous connectivity measure, including measures of human movement, and that the models can be extended to explore multiple sources of spatial structure in the data. Using Bayesian inference and simulation, the relative contribution of each spatial structure can be computed and used to generate hypotheses about the drivers of disease. These models were found to perform at least as well as existing modelling frameworks, while allowing for future extensions and multiple sources of spatial connectivity.

show abstract

“…Many area health outcomes, including psychosis, show spatial clustering [18,19], and any statistical model should incorporate this feature. To allow for spatial clustering in unobserved risk factors a common approach, known as disease mapping regression [20], is to include a spatially correlated residual term in regression models. This strategy is typically used in conjunction with Bayesian inference, and implies a spatial borrowing of strength for relatively rare health outcomes [20,21].…”

Section: Spatial Variations In Psychosis and Ecological Modelsmentioning

confidence: 99%

Psychosis Prevalence in London Neighbourhoods; a Case Study in Spatial Confounding

Congdon

2022

Preprint

View full text Add to dashboard Cite

A considerable body of research concerns spatial variations in psychosis and impacts of neighbourhood risk factors. Such research frequently adopts a disease mapping approach, with unknown spatially clustered neighbourhood influences summarised by random effects. However, added spatial random effects may show confounding with observed area predictors, especially when observed area predictors have a clear spatial pattern. In a case study application, the standard disease mapping model is compared to methods which account and adjust for spatial confounding in an analysis of psychosis prevalence in London neighbourhoods. Established area risk factors such as area deprivation, non-white ethnicity, greenspace access and social fragmentation are considered as influences on psychosis levels. The results show evidence of spatial confounding in the standard disease mapping model. Impacts expected on substantive grounds and available evidence are either nullified or reversed in direction. Inferences about excess relative psychosis risk in different small neighbourhoods are affected. It is argued that the potential for spatial confounding to affect inferences about geographic disease patterns and risk factors should be routinely considered in ecological studies of health based on disease mapping.

show abstract

Bayesian disease mapping: Past, present, and future

Cited by 20 publications

References 129 publications

Revisiting Gaussian Markov random fields and Bayesian disease mapping

Revisiting Gaussian Markov random fields and Bayesian disease mapping

A Bayesian modelling framework to quantify multiple sources of spatial variation for disease mapping

Psychosis Prevalence in London Neighbourhoods; a Case Study in Spatial Confounding

Contact Info

Product

Resources

About