Bayesian estimates of disease maps: how important are priors?

Spatiotemporal trends and ecological determinants of cardiovascular mortality among 2844 counties in mainland China, 2006-2020: a Bayesian modeling study of national mortality registries

BACKGROUND

Cardiovascular disease (CVD) is the leading cause of death in China. No previous study has reported CVD mortality at county-level, and little was known about the nonmedical ecological factors of CVD mortality at such small scale in mainland China. Understanding the spatiotemporal variations of CVD mortality and examining its nonmedical ecological factors would be of great importance to tailor local public health policies.

METHODS

By using national mortality registration data in China, this study used hierarchical spatiotemporal Bayesian model to demonstrate spatiotemporal distribution of CVD mortality in 2844 counties during 2006 to 2020 and investigate how nonmedical ecological determinants have affected CVD mortality inequities from the spatial perspectives.

RESULTS

During 2006-2020, the age-standardized mortality rate (ASMR) of CVD decreased from 284.77 per 100,000 in 2006 to 241.34 per 100,000 in 2020. Among 2844 counties, 1144 (40.22%) were hot spots counties with a higher CVD mortality risk compared to the national average and located mostly in northeast, north central, and westernmost regions; on the contrary, 1551 (54.53%) were cold spots counties and located mostly in south and southeast coastal counties. CVD mortality risk decreased from 2006 to 2020 was larger in counties where CVD mortality rate had been higher in 2006 in most of the counties, vice versa. Nationwide, nighttime light intensity (NTL) was the major influencing factor of CVD mortality, a higher NTL appeared to be negatively associated with a lower CVD mortality, with one unit increase in NTL, and the CVD mortality risk will decrease 11% (relative risk of NTL was estimated as 0.89 with 95% confidence interval of 0.83-0.94).

CONCLUSIONS

Substantial between-county discrepancies of CVD mortality distribution were observed during past 15 years in mainland China. Nonmedical ecological determinants were estimated to significantly explain the overall and local spatiotemporal patterns of this CVD mortality risk. Targeted considerations are needed to integrate primary care with clinical care through intensifying further strategies to narrow unequally distribution of CVD mortality at local scale. The approach to county-level analysis with small area models has the potential to provide novel insights into Chinese disease-specific mortality burden.

Bayesian Spatial Binary Regression for Label Fusion in Structural Neuroimaging

Alzheimer's disease is a neurodegenerative condition that accelerates cognitive decline relative to normal aging. It is of critical scientific importance to gain a better understanding of early disease mechanisms in the brain to facilitate effective, targeted therapies. The volume of the hippocampus is often used in diagnosis and monitoring of the disease. Measuring this volume via neuroimaging is difficult since each hippocampus must either be manually identified or automatically delineated, a task referred to as segmentation. Automatic hippocampal segmentation often involves mapping a previously manually segmented image to a new brain image and propagating the labels to obtain an estimate of where each hippocampus is located in the new image. A more recent approach to this problem is to propagate labels from multiple manually segmented atlases and combine the results using a process known as label fusion. To date, most label fusion algorithms employ voting procedures with voting weights assigned directly or estimated via optimization. We propose using a fully Bayesian spatial regression model for label fusion that facilitates direct incorporation of covariate information while making accessible the entire posterior distribution. Our results suggest that incorporating tissue classification (e.g, gray matter) into the label fusion procedure can greatly improve segmentation when relatively homogeneous, healthy brains are used as atlases for diseased brains. The fully Bayesian approach also produces meaningful uncertainty measures about hippocampal volumes, information which can be leveraged to detect significant, scientifically meaningful differences between healthy and diseased populations, improving the potential for early detection and tracking of the disease.

Spatio-temporal modelling of tick life-stage count data with spatially varying coefficients

There is a vast amount of geo-referenced data in many fields of study including ecological studies. Geo-referencing is usually by point referencing; that is, latitudes and longitudes or by areal referencing, which includes districts, counties, states, provinces and other administrative units. The availability of large geo-referenced datasets for modelling has necessitated the development and application of spatial statistical methods. However, spatial varying coefficients models exploring the abundance of tick counts remain limited. In this study we used data that was collected and prepared by researchers in the Department of Biological Sciences from the Old Dominion University, Virginia, USA. We modelled tick life-stage counts and abundance variability from 12 sampling locations, with 5 different habitats (numbered 1-5), three habitat types; namely: woods, edges and grass; collected monthly from May 2009 through December 2018. Spatio-temporal Poisson and spatio-temporal negative binomial (NB) count data models were fitted to the data and compared using the deviance information criteria (DIC). The NB model outperformed the Poisson models with all its DIC values being smaller than those of the Poisson model. Results showed that the covariates varied spatially across counties. There was a decreasing time (in years) effect over the study period. However, even though the time effect was decreasing over the study period, space-time interaction effects were seen to be increasing over time in York County.

Long-term air pollution and other risk factors associated with COVID-19 at the census tract level in Colorado

Previous nationwide studies have reported links between long-term concentrations of fine particulate matter (PM2.5) and COVID-19 infection and mortality rates. In order to translate these results to the state level, we use Bayesian hierarchical models to explore potential links between long-term PM2.5 concentrations and census tract-level rates of COVID-19 outcomes (infections, hospitalizations, and deaths) in Colorado. We explicitly consider how the uncertainty in PM2.5 estimates affects our results by comparing four different PM2.5 surfaces from academic and governmental organizations. After controlling for 20 census tract-level covariates, we find that our results depend heavily on the choice of PM2.5 surface. Using PM2.5 estimates from the United States EPA, we find that a 1 μg/m³ increase in long-term PM2.5 concentrations is associated with a statistically significant 26% increase in the relative risk of hospitalizations and a 34% increase in mortality. Results for all other surfaces and outcomes were not statistically significant. At the same time, we find a clear association between communities of color and COVID-19 outcomes at the Colorado census tract level that is minimally affected by the choice of PM2.5 surface. A per-interquartile range (IQR) increase in the percent of non-African American people of color was associated with a 31%, 43%, and 56% increase in the relative risk of infection, hospitalization, and mortality respectively, while a per-IQR increase in the proportion of non-Hispanic African Americans was associated with a 4% and 7% increase in the relative risk of infections and hospitalizations. The current disagreement among the different PM2.5 estimates is a key factor limiting our ability to link environmental exposures and health outcomes at the census tract level. These results have strong implications for the implementation of an equitable public health response during the crisis and suggest targeted areas for additional air monitoring in Colorado.

Inequalities of visceral leishmaniasis case-fatality in Brazil: A multilevel modeling considering space, time, individual and contextual factors

Background

In Brazil, case-fatality from visceral leishmaniasis (VL) is high and characterized by wide differences between the various political-economic units, the federated units (FUs). This study was designed to investigate the association between factors at the both FU and individual levels with the risk of dying from VL, after analysing the temporal trend and the spatial dependency for VL case-fatality.

Methodology

The analysis was based on individual and aggregated data of the Reportable Disease Information System-SINAN (Brazilian Ministry of Health). The temporal and spatial distributions of the VL case-fatality between 2007 and 2017 (27 FUs as unit of analysis) were considered together with the individual characteristics and many other variables at the FU level (socioeconomic, demographic, access to health and epidemiological indicators) in a mixed effects models or multilevel modeling, assuming a binomial outcome distribution (death from VL).

Findings

A linear increasing temporal tendency (4%/year) for VL case-fatality was observed between 2007 and 2017. There was no similarity between the case-fatality rates of neighboring FUs (non-significant spatial term), although these rates were heterogeneous in this spatial scale of analysis. In addition to the known individual risk factors age, female gender, disease’s severity, bacterial co-infection and disease duration, low level schooling and unavailability of emergency beds and health professionals (the last two only in univariate analysis) were identified as possibly related to VL death risk. Lower VL incidence was also associated to VL case-fatality, suggesting that unfamiliarity with the disease may delay appropriate medical management: VL patients with fatal outcome were notified and had VL treatment started 6 and 3 days later, respectively, in relation to VL cured patients. Access to garbage collection, marker of social and economic development, seems to be protective against the risk of dying from VL. Part of the observed VL case-fatality variability in Brazil could not be explained by the studied variables, suggesting that factors linked to the intra FU environment may be involved.

Conclusions

This study aimed to identify epidemiological conditions and others related to access to the health system possibly linked to VL case-fatality, pointing out new prognostic determinants subject to intervention.

Visceral leishmaniasis (VL) is a potentially fatal disease if not diagnosed and treated promptly. The VL case-fatality in Brazil is the highest rate in the world, reaching an average of 7% and in some regions, more than 15%. In the last years, some improvements in the VL approach have been reached in Brazil, such as the widespread use of rapid diagnostic tests and liposomal amphotericin B for treatment of selected high risk of death cases. Despite these interventions, increase in case-fatality rates were observed. In this study we explored the factors related to the case-fatality from VL using a mixed modeling that encompasses different intervening factors such as time/spatial trends and factors linked to the individual and socio-economic indicators. For the first time, factors unrelated to the patients’ clinical condition emerge as possibly related to VL case-fatality, such as low educational level, unavailability of emergency beds and health professionals, suggesting the harmful influence of conditions of limited access to health services. In addition to these significant effects observed in the spatial scale of analysis, this study points to the influence of contextual factors linked to each geopolitical unit. The determinants of death among VL cases may differ according to the region, which requires specific actions planned locally, including increased access to health system qualified to recognize and properly treat VL.

Evaluating the informativeness of the Besag-York-Mollié CAR model

The use of the conditional autoregressive framework proposed by Besag, York, and Mollié (1991; BYM) is ubiquitous in Bayesian disease mapping and spatial epidemiology. While it is understood that Bayesian inference is based on a combination of the information contained in the data and the information contributed by the model, quantifying the contribution of the model relative to the information in the data is often non-trivial. Here, we provide a measure of the contribution of the BYM framework by first considering the simple Poisson-gamma setting in which quantifying the prior's contribution is quite clear. We then propose a relationship between gamma and lognormal priors that we then extend to cover the framework proposed by BYM. Following a brief simulation study in which we illustrate the accuracy of our lognormal approximation of the gamma prior, we analyze a dataset comprised of county-level heart disease-related death data across the United States. In addition to demonstrating the potential for the BYM framework to correspond to a highly informative prior specification, we also illustrate the sensitivity of death rate estimates to changes in the informativeness of the BYM framework.

Modeling epidemic spread in transportation networks: A review

The emergence of novel infectious diseases has become a serious global problem. Convenient transportation networks lead to rapid mobilization in the context of globalization, which is an important factor underlying the rapid spread of infectious diseases. Transportation systems can cause the transmission of viruses during the epidemic period, but they also support the reopening of economies after the epidemic. Understanding the mechanism of the impact of mobility on the spread of infectious diseases is thus important, as is establishing the risk model of the spread of infectious diseases in transportation networks. In this study, the basic structure and application of various epidemic spread models are reviewed, including mathematical models, statistical models, network-based models, and simulation models. The advantages and limitations of model applications within transportation systems are analyzed, including dynamic characteristics of epidemic transmission and decision supports for management and control. Lastly, research trends and prospects are discussed. It is suggested that there is a need for more in-depth research to examine the mutual feedback mechanism of epidemics and individual behavior, as well as the proposal and evaluation of intervention measures. The findings in this study can help evaluate disease intervention strategies, provide decision supports for transport policy during the epidemic period, and ameliorate the deficiencies of the existing system.

•

Reviewed epidemic spread models and their applications in transportation networks.

•

Analyzed the advantages and limitations of epidemic spread model applications in transportation systems.

•

Summarized the emerging modeling requirements brought by the COVID-19 pandemic.

•

Proposed research trends and prospects for epidemic spread modeling in transportation networks.

Use of geographical information systems in multiple sclerosis research: A systematic scoping review

INTRODUCTION

Geographical information system (GIS) and spatial analysis have an emerging role in the understanding and management of health-related outcomes. However, there is a knowledge gap about the extent to which GIS has supported multiple Sclerosis (MS) research. Therefore, this review aimed to explore the types of GIS applications and the complexity of their visualisation in MS research.

METHODS

A systematic scoping review was conducted based on York's five-stage framework. PubMed, Scopus and Web of Science were searched for relevant studies published between 2000 and 2020 using a comprehensive search strategy based on the main concepts related to GIS and MS. Grounded, inductive analysis was conducted to organize studies into meaningful application areas. Further, we developed a tool to assess the visualisation complexity of the selected papers.

RESULTS

Of 3,723 identified unique citations, 42 papers met our inclusion criteria for the final review. One or more of the following types of GIS applications were reported by these studies: (a) thematic mapping (37 papers); (b) spatial cluster detection (16 papers); (c) risk factors detection (16 papers); and (d) health access and planning (two papers). In the majority of studies (88%), the score of visualisation complexity was relatively low: three or less from the range of zero to six.

CONCLUSIONS

Although the number of studies using GIS techniques has dramatically increased in the last decade, the use of GIS in the areas of MS access and planning is still under-researched. Additionally, the capacity of GIS in visualising complex nature of MS care system is not yet fully investigated.

Spatial-temporal patterns of malaria incidence in Uganda using HMIS data from 2015 to 2019

BACKGROUND

As global progress to reduce malaria transmission continues, it is increasingly important to track changes in malaria incidence rather than prevalence. Risk estimates for Africa have largely underutilized available health management information systems (HMIS) data to monitor trends. This study uses national HMIS data, together with environmental and geographical data, to assess spatial-temporal patterns of malaria incidence at facility catchment level in Uganda, over a recent 5-year period.

METHODS

Data reported by 3446 health facilities in Uganda, between July 2015 and September 2019, was analysed. To assess the geographic accessibility of the health facilities network, AccessMod was employed to determine a three-hour cost-distance catchment around each facility. Using confirmed malaria cases and total catchment population by facility, an ecological Bayesian conditional autoregressive spatial-temporal Poisson model was fitted to generate monthly posterior incidence rate estimates, adjusted for caregiver education, rainfall, land surface temperature, night-time light (an indicator of urbanicity), and vegetation index.

RESULTS

An estimated 38.8 million (95% Credible Interval [CI]: 37.9-40.9) confirmed cases of malaria occurred over the period, with a national mean monthly incidence rate of 20.4 (95% CI: 19.9-21.5) cases per 1000, ranging from 8.9 (95% CI: 8.7-9.4) to 36.6 (95% CI: 35.7-38.5) across the study period. Strong seasonality was observed, with June-July experiencing highest peaks and February-March the lowest peaks. There was also considerable geographic heterogeneity in incidence, with health facility catchment relative risk during peak transmission months ranging from 0 to 50.5 (95% CI: 49.0-50.8) times higher than national average. Both districts and health facility catchments showed significant positive spatial autocorrelation; health facility catchments had global Moran's I = 0.3 (p < 0.001) and districts Moran's I = 0.4 (p < 0.001). Notably, significant clusters of high-risk health facility catchments were concentrated in Acholi, West Nile, Karamoja, and East Central - Busoga regions.

CONCLUSION

Findings showed clear countrywide spatial-temporal patterns with clustering of malaria risk across districts and health facility catchments within high risk regions, which can facilitate targeting of interventions to those areas at highest risk. Moreover, despite high and perennial transmission, seasonality for malaria incidence highlights the potential for optimal and timely implementation of targeted interventions.

Bayesian spatial modelling of early childhood development in Australian regions

Background

Children’s early development plays a vital role for maintaining healthy lives and influences future outcomes. It is also heavily affected by community factors which vary geographically. Direct methods do not provide a comprehensive picture of this variation, especially for areas with sparse populations and low data coverage. In the context of Australia, the Australian Early Development Census (AEDC) provides a measure of early child development upon school entry. There are two primary aims of this study: (i) provide improved prevalence estimates of children who are considered as developmentally vulnerable in regions across Australia; (ii) ascertain how social-economic disadvantage partly explains the spatial variation.

Methods

We used Bayesian spatial hierarchical models with the Socio-economic Indexes for Areas (SEIFA) as a covariate to provide improved estimates of all 335 SA3 regions in Australia. The study included 308,953 children involved in the 2018 AEDC where 21.7% of them were considered to be developmentally vulnerable in at least one domain. There are five domains of developmental vulnerability—physical health and wellbeing; social competence; emotional maturity; language and cognitive skills; and communication and general knowledge.

Results

There are significant improvements in estimation of the prevalence of developmental vulnerability through incorporating the socio-economic disadvantage in an area. These improvements persist in all five domains—the largest improvements occurred in the Language and Cognitive Skills domain. In addition, our results reveal that there is an important geographical dimension to developmental vulnerability in Australia.

Conclusion

Sparsely populated areas in sample surveys lead to unreliable direct estimates of the relatively small prevalence of child vulnerability. Bayesian spatial modelling can account for the spatial patterns in childhood vulnerability while including the impact of socio-economic disadvantage on geographic variation. Further investigation, using a broader range of covariates, could shed more light on explaining this spatial variation.

Spatiotemporal Analysis of Influenza in China, 2005-2018

Influenza is a major cause of morbidity and mortality worldwide, as well as in China. Knowledge of the spatial and temporal characteristics of influenza is important in evaluating and developing disease control programs. This study aims to describe an accurate spatiotemporal pattern of influenza at the prefecture level and explore the risk factors associated with influenza incidence risk in mainland China from 2005 to 2018. The incidence data of influenza were obtained from the Chinese Notifiable Infectious Disease Reporting System (CNIDRS). The Besag York Mollié (BYM) model was extended to include temporal and space-time interaction terms. The parameters for this extended Bayesian spatiotemporal model were estimated through integrated nested Laplace approximations (INLA) using the package R-INLA in R. A total of 702,226 influenza cases were reported in mainland China in CNIDRS from 2005–2018. The yearly reported incidence rate of influenza increased 15.6 times over the study period, from 3.51 in 2005 to 55.09 in 2008 per 100,000 populations. The temporal term in the spatiotemporal model showed that much of the increase occurred during the last 3 years of the study period. The risk factor analysis showed that the decreased number of influenza vaccines for sale, the new update of the influenza surveillance protocol, the increase in the rate of influenza A (H1N1)pdm09 among all processed specimens from influenza-like illness (ILI) patients, and the increase in the latitude and longitude of geographic location were associated with an increase in the influenza incidence risk. After the adjusting for fixed covariate effects and time random effects, the map of the spatial structured term shows that high-risk areas clustered in the central part of China and the lowest-risk areas in the east and west. Large space-time variations in influenza have been found since 2009. In conclusion, an increasing trend of influenza was observed from 2005 to 2018. The insufficient flu vaccine supplements, the newly emerging influenza A (H1N1)pdm09 and expansion of influenza surveillance efforts might be the major causes of the dramatic changes in outbreak and spatio-temporal epidemic patterns. Clusters of prefectures with high relative risks of influenza were identified in the central part of China. Future research with more risk factors at both national and local levels is necessary to explain the changing spatiotemporal patterns of influenza in China.

Spatial heterogeneity of KRAS mutations in colorectal cancers in northern France

Background

Somatic mutations in the KRAS gene are the most common oncogenic mutations found in human cancers. However, no clinical features have been linked to KRAS mutations in colorectal cancer [CRC].

Purpose

In this study, we attempted to identify the potential geographical population clusters of KRAS mutations in CRC patients in northern France.

Patients and methods

All patients with CRC who were identified to have KRAS mutations between 2008 and 2014 at the Regional Molecular Biology Platform at Lille University Hospital were included. 2,486 patients underwent a KRAS status available, with 40.9% of CRC with KRAS mutations in northern France. We retrospectively collected demographic and geographic data from these patients. The proportions of KRAS mutation were smoothed to take into account the variability related to low frequencies and spatial autocorrelation. Geographical clusters were searched using spatial scan statistical models.

Results

A mutation at KRAS codon 12 or 13 was found in 1,018 patients [40.9%]. We report 5 clusters of over-incidence but only one elongated cluster that was statistically significant [Cluster 1; proportion of KRAS mutation among CRC: 0.4570; RR=1.29; P=0.0314]. We made an ecological study which did not highlight a significant association between KRAS mutations and the distance to the Closest Waste Incineration Plant, and between KRAS mutations and The French Ecological Deprivation Index but few socio-economic and environmental data were available.

Conclusion

There was a spatial heterogeneity and a greater frequency of KRAS mutations in some areas close to major highways and big cities in northern France. These data demand deeper epidemiological investigations to identify environmental factors such as air pollution as key factors in the occurrence of KRAS mutations.

Bayesian hierarchical spatial models: Implementing the Besag York Mollié model in stan

This report presents a new implementation of the Besag-York-Mollié (BYM) model in Stan, a probabilistic programming platform which does full Bayesian inference using Hamiltonian Monte Carlo (HMC). We review the spatial auto-correlation models used for areal data and disease risk mapping, and describe the corresponding Stan implementations. We also present a case study using Stan to fit a BYM model for motor vehicle crashes injuring school-age pedestrians in New York City from 2005 to 2014 localized to census tracts. Stan efficiently fit our multivariable BYM model having a large number of observations (n=2095 census tracts) with small outcome counts < 10 in the majority of tracts. Our findings reinforced that neighborhood income and social fragmentation are significant correlates of school-age pedestrian injuries. We also observed that nationally-available census tract estimates of commuting methods may serve as a useful indicator of underlying pedestrian densities.

Exploring socio-demographic-and geographical-variations in prevalence of diabetes and hypertension in Bangladesh: Bayesian spatial analysis of national health survey data

Bangladesh has been experiencing an epidemiological transition from communicable diseases to non-communicable disease (NCDs), with a rapid increase in the NCD related morbidity and mortality in the last decade. Hypertension and diabetes are two important risk factors of NCDs that significantly increase the burden of cardiovascular diseases and risk of death. While the prevalence of people with both hypertension and diabetes has been increasing dramatically over time, it is essential to identify relatively more prevalent socio-demographic groups and geographical regions (local administrative districts) to reduce the NCDs related deaths in an urgent basis. This study focused on examining the association of socio-demographic factors with both hypertension and diabetes and exploring the regional variations in their prevalence using nationally representative survey data on adult population of age over 35 years. Bayesian spatial analysis was performed for both hypertension and diabetes data separately by fitting a model, that accounts for spatial variations, using integrated nested laplace approximation. The area-specific prevalence was then estimated as weighted average of the corresponding individual level predicted probabilities of being diseased derived from the fitted model, with weight from the individual level sampling weight. Finally, the estimated area-specific prevalence estimates were sketched in country-map to explore regional variations and identify regions with relatively higher prevalence. The results revealed that people of older age, higher education, better socio-economic condition, higher BMI are at greater risk of having hypertension and diabetes. Significant regional variations were observed with prevalence for hypertension ranges between 10% and 35% and for diabetes between 6% and 19% while their national prevalence were reported as 24% and 11%, respectively. The western regions of the country including middle capital city were found to be relatively more prevalent for hypertension while the middle-east and south-east regions were observed to be more prevalent for diabetes. The capital Dhaka region was observed as the most prevalent for both diabetes and hypertension. Details explanations of the findings and evidence based policy implications were discussed.

Spatial analysis of macro-level bicycle crashes using the class of conditional autoregressive models

The objective of this study was to investigate the relationship between bicycle crash frequency and their contributing factors at the census block group level in Florida, USA. Crashes aggregated over the census block groups tend to be clustered (i.e., spatially dependent) rather than randomly distributed. To account for the effect of spatial dependence across the census block groups, the class of conditional autoregressive (CAR) models were employed within the hierarchical Bayesian framework. Based on four years (2011-2014) of crash data, total and fatal-and-severe injury bicycle crash frequencies were modeled as a function of a large number of variables representing demographic and socio-economic characteristics, roadway infrastructure and traffic characteristics, and bicycle activity characteristics. This study explored and compared the performance of two CAR models, namely the Besag's model and the Leroux's model, in crash prediction. The Besag's models, which differ from the Leroux's models by the structure of how spatial autocorrelation are specified in the models, were found to fit the data better. A 95% Bayesian credible interval was selected to identify the variables that had credible impact on bicycle crashes. A total of 21 variables were found to be credible in the total crash model, while 18 variables were found to be credible in the fatal-and-severe injury crash model. Population, daily vehicle miles traveled, age cohorts, household automobile ownership, density of urban roads by functional class, bicycle trip miles, and bicycle trip intensity had positive effects in both the total and fatal-and-severe crash models. Educational attainment variables, truck percentage, and density of rural roads by functional class were found to be negatively associated with both total and fatal-and-severe bicycle crash frequencies.

An intuitive Bayesian spatial model for disease mapping that accounts for scaling

In recent years, disease mapping studies have become a routine application within geographical epidemiology and are typically analysed within a Bayesian hierarchical model formulation. A variety of model formulations for the latent level have been proposed but all come with inherent issues. In the classical BYM (Besag, York and Mollié) model, the spatially structured component cannot be seen independently from the unstructured component. This makes prior definitions for the hyperparameters of the two random effects challenging. There are alternative model formulations that address this confounding; however, the issue on how to choose interpretable hyperpriors is still unsolved. Here, we discuss a recently proposed parameterisation of the BYM model that leads to improved parameter control as the hyperparameters can be seen independently from each other. Furthermore, the need for a scaled spatial component is addressed, which facilitates assignment of interpretable hyperpriors and make these transferable between spatial applications with different graph structures. The hyperparameters themselves are used to define flexible extensions of simple base models. Consequently, penalised complexity priors for these parameters can be derived based on the information-theoretic distance from the flexible model to the base model, giving priors with clear interpretation. We provide implementation details for the new model formulation which preserve sparsity properties, and we investigate systematically the model performance and compare it to existing parameterisations. Through a simulation study, we show that the new model performs well, both showing good learning abilities and good shrinkage behaviour. In terms of model choice criteria, the proposed model performs at least equally well as existing parameterisations, but only the new formulation offers parameters that are interpretable and hyperpriors that have a clear meaning.

Spatially varying auto-regressive models for prediction of new human immunodeficiency virus diagnoses

In demand of predicting new HIV diagnosis rates based on publicly available HIV data that is abundant in space but has few points in time, we propose a class of spatially varying autoregressive (SVAR) models compounded with conditional autoregressive (CAR) spatial correlation structures. We then propose to use the copula approach and a flexible CAR formulation to model the dependence between adjacent counties. These models allow for spatial and temporal correlation as well as space-time interactions and are naturally suitable for predicting HIV cases and other spatio-temporal disease data that feature a similar data structure. We apply the proposed models to HIV data over Florida, California and New England states and compare them to a range of linear mixed models that have been recently popular for modeling spatio-temporal disease data. The results show that for such data our proposed models outperform the others in terms of prediction.

Bayesian spatial models of the association between interpersonal violence, animal abuse and social vulnerability in São Paulo, Brazil

Animal abuse adversely affects animal health and welfare and has been associated with interpersonal violence in studies of individuals. However, if that association also depends on sociocultural contexts and can be detected on a geographic scale, a wider source of data can be used to identify risk areas to support the surveillance of both types of violence. In this study, we evaluated the association between interpersonal violence notifications, animal abuse notifications and an index of social vulnerability in São Paulo City, on a geographic scale, using Bayesian spatial models. The social vulnerability index was a risk factor for the number of interpersonal violence notifications and presented a dose-response pattern. The number of animal abuse notifications was also a risk factor for the number of interpersonal violence notifications, even after controlling for the social vulnerability index. The incorporation of spatial effects produced marked improvements in model performance metrics and allowed the identification of excess risk clusters. Geographical data on notifications on either animal abuse or interpersonal violence should be considered incitement for investigations and interventions of both types of violence. We suggest that notifications of animal abuse be based on an explicit definition and classification, as well as on objective measurements that allow a better understanding of the species and type of abuse involved, the animal health consequences, and the context in which they occurred.

Using additive and coupled spatiotemporal SPDE models: a flexible illustration for predicting occurrence of Culicoides species

This paper formulates and compares a general class of spatiotemporal models for univariate space-time geostatistical data. The implementation of stochastic partial differential equation (SPDE) approach combined with integrated nested Laplace approximation into the R-INLA package makes it computationally feasible to use spatiotemporal models. However, the impact of specifying models with and without space-time interaction is unclear. We formulate an extensive class of additive and coupled spatiotemporal SPDE models and investigate the distinction between them by (1) Extending their temporal effect, allowing a random walk process in time, (2) varying the spatial correlation function and (3) running a simulation study to assess the effect of misspecifying the spatial and temporal models, and to assess the generalizability of our results to a higher number of locations. Our methods are illustrated with Culicoides data from Belgium. The Bayesian spatial predictions showed that the highest prevalence of Culicoides species was found in the Northeastern and central parts of Belgium during summer.

A tall order: Small area mapping and modelling of adult height among Swiss male conscripts

Adult height reflects an individual's socio-economic background and offers insights into the well-being of populations. Height is linked to various health outcomes such as morbidity and mortality and has consequences on the societal level. The aim of this study was to describe small-area variation of height and associated factors among young men in Switzerland. Data from 175,916 conscripts (aged between 18.50 and 20.50 years) was collected between 2005 and 2011, which represented approximately 90% of the corresponding birth cohorts. These were analysed using Gaussian hierarchical models in a Bayesian framework to investigate the spatial pattern of mean height across postcodes. The models varied both in random effects and degree of adjustment (professional status, area-based socioeconomic position, and language region). We found a strong spatial structure for mean height across postcodes. The range of height differences between mean postcode level estimates was 3.40cm according to the best fitting model, with the shorter conscripts coming from the Italian and French speaking parts of Switzerland. There were positive socioeconomic gradients in height at both individual and area-based levels. Spatial patterns for height persisted after adjustment for individual factors, but not when language region was included. Socio-economic position and cultural/natural boundaries such as language borders and mountain passes are shaping patterns of height for Swiss conscripts. Small area mapping of height contributes to the understanding of its cofactors.

Bayesian spatial analysis of the surgical incidence rate of human cystic echinococcosis in north-eastern Iran

BACKGROUND

Cystic echinococcosis (CE) is a zoonotic disease that presents a public health challenge and a socioeconomic burden on developing areas in the Middle East. This study used spatial methods to assess the distribution of surgically managed CE cases in an endemic region of north-eastern Iran.

METHODS

For the years 2001-2007, a case series of all 446 patients that were surgically treated for CE in a referral hospital in north-eastern Iran was evaluated. Patients seen at the referral hospital represent 35 counties in three provinces (Razavi Khorasan, North Khorasan, and South Khorasan). A Besag, York and Mollie (BYM) spatial model was used to produce smoothed standardized incidence ratios (SIRs) for surgically managed cases of CE for the 35 counties represented in this study.

RESULTS

Out of 446 surgically managed patients, 54% were male. County-level crude incidence rates ranged from 0 to 3.27 cases per 100,000 population. The highest smoothed SIR (3.46) was for Sarakhs County in the province of Razavi Khorasan, while the lowest smoothed SIR (0.05) was for Birjand County, located in the province of South Khorasan.

CONCLUSION

SIRs for CE were highest for the province of Razavi Khorasan, which has large ranching and agricultural industries. Additional studies are needed to better evaluate the role of climate, land cover, and livestock rearing on local Echinococcus granulosus transmission in Iran.

A full Bayes before-after study accounting for temporal and spatial effects: Evaluating the safety impact of new signal installations

Recently, important advances in road safety statistics have been brought about by methods able to address issues other than the choice of the best error structure for modeling crash data. In particular, accounting for spatial and temporal interdependence, i.e., the notion that the collision occurrence of a site or unit times depend on those of others, has become an important issue that needs further research. Overall, autoregressive models can be used for this purpose as they can specify that the output variable depends on its own previous values and on a stochastic term. Spatial effects have been investigated and applied mostly in the context of developing safety performance functions (SPFs) to relate crash occurrence to highway characteristics. Hence, there is a need for studies that attempt to estimate the effectiveness of safety countermeasures by including the spatial interdependence of road sites within the context of an observational before-after (BA) study. Moreover, the combination of temporal dynamics and spatial effects on crash frequency has not been explored in depth for SPF development. Therefore, the main goal of this research was to carry out a BA study accounting for spatial effects and temporal dynamics in evaluating the effectiveness of a road safety treatment. The countermeasure analyzed was the installation of traffic signals at unsignalized urban/suburban intersections in British Columbia (Canada). The full Bayes approach was selected as the statistical framework to develop the models. The results demonstrated that zone variation was a major component of total crash variability and that spatial effects were alleviated by clustering intersections together. Finally, the methodology used also allowed estimation of the treatment's effectiveness in the form of crash modification factors and functions with time trends.

Impact of large industrial emission sources on mortality and morbidity in Chile: A small-areas study

Chile suffers significant pollution from large industrial emitters associated with the mining, metal processing, paper production, and energy industries. The aim of this research was to determine whether the presence of large industrial facilities (i.e. coal- and oil-fired power plants, pulp and paper mills, mining facilities, and smelters) affects mortality and morbidity rates in Chile. For this, we conducted an ecological study that used Chilean communes as small-area observation units to assess mortality and morbidity. Public databases provided information on large pollution sources relevant to Chile. The large sources studied were oil- and coal-fired power plants, copper smelters, pulp and paper mills, and large mining facilities. Large sources were filtered by first year of production, type of process, and size. Mortality and morbidity data were acquired from public national databases, with morbidity being estimated from hospitalization records. Cause-specific rates were calculated for the main outcomes: cardiovascular, respiratory, cancer; and other more specific health outcomes. The impact of the large pollution sources was estimated using Bayesian models that included spatial correlation, overdispersion, and other covariates. Large and significant increases in health risks (around 20%-100%) were found for communes with power plants and smelters for total, cardiovascular, respiratory, all-cancer, and lung cancer mortality. Higher hospitalization rates for cardiovascular disease, respiratory disease, cancer, and pneumonia (20-100%) were also found for communes with power plants and smelters. The impacts were larger for men than women in terms of both mortality and hospitalizations. The impacts were also larger when the sources were analyzed as continuous (production volume) rather than dichotomous (presence/absence) variables. In conclusion, significantly higher rates of total cardiovascular, respiratory, all-cancer and lung cancer mortality and cardiovascular, respiratory, cancer and pneumonia hospitalizations were observed in communes with power plants and smelters.

Spatial mixture multiscale modeling for aggregated health data

One of the main goals in spatial epidemiology is to study the geographical pattern of disease risks. For such purpose, the convolution model composed of correlated and uncorrelated components is often used. However, one of the two components could be predominant in some regions. To investigate the predominance of the correlated or uncorrelated component for multiple scale data, we propose four different spatial mixture multiscale models by mixing spatially varying probability weights of correlated (CH) and uncorrelated heterogeneities (UH). The first model assumes that there is no linkage between the different scales and, hence, we consider independent mixture convolution models at each scale. The second model introduces linkage between finer and coarser scales via a shared uncorrelated component of the mixture convolution model. The third model is similar to the second model but the linkage between the scales is introduced through the correlated component. Finally, the fourth model accommodates for a scale effect by sharing both CH and UH simultaneously. We applied these models to real and simulated data, and found that the fourth model is the best model followed by the second model.

Finding big shots: small-area mapping and spatial modelling of obesity among Swiss male conscripts

BACKGROUND

In Switzerland, as in other developed countries, the prevalence of overweight and obesity has increased substantially since the early 1990s. Most of the analyses so far have been based on sporadic surveys or self-reported data and did not offer potential for small-area analyses. The goal of this study was to investigate spatial variation and determinants of obesity among young Swiss men using recent conscription data.

METHODS

A complete, anonymized dataset of conscription records for the 2010-2012 period were provided by Swiss Armed Forces. We used a series of Bayesian hierarchical logistic regression models to investigate the spatial pattern of obesity across 3,187 postcodes, varying them by type of random effects (spatially unstructured and structured), level of adjustment by individual (age and professional status) and area-based [urbanicity and index of socio-economic position (SEP)] characteristics.

RESULTS

The analysed dataset consisted of 100,919 conscripts, out of which 5,892 (5.8 %) were obese. Crude obesity prevalence increased with age among conscripts of lower individual and area-based SEP and varied greatly over postcodes. Best model's estimates of adjusted odds ratios of obesity on postcode level ranged from 0.61 to 1.93 and showed a strong spatial pattern of obesity risk across the country. Odds ratios above 1 concentrated in central and north Switzerland. Smaller pockets of elevated obesity risk also emerged around cities of Geneva, Fribourg and Lausanne. Lower estimates were observed in North-East and East as well as south of the Alps. Importantly, small regional outliers were observed and patterning did not follow administrative boundaries. Similarly as with crude obesity prevalence, the best fitting model confirmed increasing risk of obesity with age and among conscripts of lower professional status. The risk decreased with higher area-based SEP and, to a lesser degree - in rural areas.

CONCLUSION

In Switzerland, there is a substantial spatial variation in obesity risk among young Swiss men. Small-area estimates of obesity risk derived from conscripts records contribute to its understanding and could be used to design further studies and interventions.

Análise espaço-temporal da leishmaniose visceral no estado do Maranhão, Brasil

Resumo Este estudo analisou a distribuição espaço-temporal dos casos de leishmaniose visceral (LV) no estado do Maranhão, no período de 2000 a 2009. A partir do número de casos notificados, foram elaborados mapas temáticos para demonstrar a evolução da distribuição geográfica da doença no estado. Utilizou-se o método MCMC para estimação dos parâmetros do modelo bayesiano espaço-temporal para a identificação das áreas de risco. De 2000 a 2009, foram notificados 5.389 casos de leishmaniose visceral, distribuídos em todas as 18 Unidades Regionais de Saúde do estado, com as maiores incidências em: Caxias, Imperatriz, Presidente Dutra e Chapadinha. As Unidades Regionais de Saúde com maiores riscos relativos por biênio foram: Caxias e Barra do Corda (2000-2001), Imperatriz e Presidente Dutra (2002-2003), Imperatriz e Caxias (2004-2005), Presidente Dutra e Codó (2006-2007), e Imperatriz e Caxias (2008-2009). Houve uma considerável expansão geográfica da LV no Maranhão, sendo necessária a adoção de medidas mais eficazes de prevenção e controle da doença no estado.

Geographical distribution of alcohol-attributable mortality in Chile: a Bayesian spatial analysis

OBJECTIVE

To describe the distribution of alcohol-attributable mortality (AAM) at the local level (345 municipalities) in Chile, including fully and partially attributable causes in 2009.

METHODS

AAM was estimated for the population 15years of age and older using per capita alcohol consumption combined with survey estimates. The effect of alcohol on each cause of death was extracted from the published scientific literature. We used Bayesian hierarchical models to smooth the Standardized Mortality Ratio for each municipality for six groups of causes related to alcohol consumption (total, neuro-psychiatric, cardiovascular, cancer, injuries and other causes).

RESULTS

The percentage of municipalities with high risk for any group of causes in each region ranges from 0% to 87.0%. Municipalities with high risk were concentrated in south-central and southern Chile for all groups of causes related to alcohol.

CONCLUSIONS

AAM risk shows marked geographic concentrations, mainly in south-central and southern regions of Chile. This combination of methods for small-area estimates of AAM is a powerful tool to identify high risk regions and associated factors, and may be used to inform local policies and programs.

Trends of breast cancer incidence in Iran during 2004-2008: a Bayesian space-time model

BACKGROUND

Breast cancer is the most frequently diagnosed cancer in women and estimating its relative risks and trends of incidence at the area-level is helpful for health policy makers. However, traditional methods of estimation which do not take spatial heterogeneity into account suffer from drawbacks and their results may be misleading, as the estimated maps of incidence vary dramatically in neighboring areas. Spatial methods have been proposed to overcome drawbacks of traditional methods by including spatial sources of variation in the model to produce smoother maps.

MATERIALS AND METHODS

In this study we analyzed the breast cancer data in Iran during 2004-2008. We used a method proposed to cover spatial and temporal effects simultaneously and their interactions to study trends of breast cancer incidence in Iran.

RESULTS

The results agree with previous studies but provide new information about two main issues regarding the trend of breast cancer in provinces of Iran. First, this model discovered provinces with high relative risks of breast cancer during the 5 years of the study. Second, new information was provided with respect to overall trend trends o. East-Azerbaijan, Golestan, North-Khorasan, and Khorasan-Razavi had the highest increases in rates of breast cancer incidence whilst Tehran, Isfahan, and Yazd had the highest incidence rates during 2004-2008.

CONCLUSIONS

Using spatial methods can provide more accurate and detailed information about the incidence or prevalence of a disease. These models can specify provinces with different health priorities in terms of needs for therapy and drugs or demands for efficient education, screening, and preventive policy into action.

Mapping end-stage renal disease (ESRD): spatial variations on small area level in northern France, and association with deprivation

BACKGROUND

Strong geographic variations in the incidence of end-stage renal disease (ESRD) are observed in developed countries. The reasons for these variations are unknown. They may reflect regional inequalities in the population's sociodemographic characteristics, related diseases, or medical practice patterns. In France, at the district level, the highest incidence rates have been found in the Nord-Pas-de-Calais region. This area, with a high population density and homogeneous healthcare provision, represents a geographic situation which is quite suitable for the study, over small areas, of spatial disparities in the incidence of ESRD, together with their correlation with a deprivation index and other risk factors.

METHODS

The Renal Epidemiology and Information Network is a national registry, which lists all ESRD patients in France. All cases included in the Nord-Pas-de-Calais registry between 2005 and 2011 were extracted. Adjusted and smoothed standardized incidence ratio (SIR) was calculated for each of the 170 cantons, thanks to a hierarchical Bayesian model. The correlation between ESRD incidence and deprivation was assessed using the quintiles of Townsend index. Relative risk (RR) and credible intervals (CI) were estimated for each quintile.

RESULTS

Significant spatial disparities in ESRD incidence were found within the Nord-Pas-de-Calais region. The sex- and age-adjusted, smoothed SIRs varied from 0.66 to 1.64. Although no correlation is found with diabetic or vascular nephropathy, the smoothed SIRs are correlated with the Townsend index (RR: 1.18, 95% CI [1.00-1.34] for Q2; 1.28, 95% CI [1.11-1.47] for Q3; 1.30, 95% CI [1.14-1.51] for Q4; 1.44, 95% CI [1.32-1.74] for Q5).

CONCLUSION

For the first time at this aggregation level in France, this study reveals significant geographic differences in ESRD incidence. Unlike the time of renal replacement care, deprivation is certainly a determinant in this phenomenon. This association is probably independent of the patients' financial ability to gain access to healthcare.

Neighborhood geographic disparities in heart attack and stroke mortality: comparison of global and local modeling approaches

This study investigated neighborhood geographic disparities in myocardial infarction (MI) and stroke mortality risks in middle Tennessee and identified determinants of observed disparities. Descriptive and spatial analyses were performed on MI and stroke mortality data covering the time period 1999-2007. Besag, York and Molliè (BYM) model was used to investigate spatial patterns. Global (BYM) and local models [Poisson Geographically Weighted Generalized Linear Models (GWGLM)] were used to investigate determinants of the identified spatial patterns. Significant (p<0.05) differences in mortality risks by sex, race, age and education were observed. Rural census tracts (CT) and those with higher proportions of the older populations were associated with high MI and stroke mortality risks. Additionally, CTs with high proportions of widows had significantly higher mortality risks for stroke. There was evidence of geographical variability of all regression coefficients implying that local models complement the findings of the global models and provide useful information to guide local and regional disease control decisions and resource allocation. Identification of high risk CTs is essential for targeting resources and will aid the development of more needs-based prevention programs.

Geographic variation of failure-to-rescue in public acute hospitals in New South Wales, Australia

Despite the wide acceptance of Failure-to-Rescue (FTR) as a patient safety indicator (defined as the deaths among surgical patients with treatable complications), no study has explored the geographic variation of FTR in a large health jurisdiction. Our study aimed to explore the spatiotemporal variations of FTR rates across New South Wales (NSW), Australia. We conducted a population-based study using all admitted surgical patients in public acute hospitals during 2002-2009 in NSW, Australia. We developed a spatiotemporal Poisson model using Integrated Nested Laplace Approximation (INLA) methods in a Bayesian framework to obtain area-specific adjusted relative risk. Local Government Area (LGA) was chosen as the areal unit. LGA-aggregated covariates included age, gender, socio-economic and remoteness index scores, distance between patient residential postcode and the treating hospital, and a quadratic time trend. We studied 4,285,494 elective surgical admissions in 82 acute public hospitals over eight years in NSW. Around 14% of patients who developed at least one of the six FTR-related complications (58,590) died during hospitalization. Of 153 LGAs, patients who lived in 31 LGAs, accommodating 48% of NSW patients at risk, were exposed to an excessive adjusted FTR risk (10% to 50%) compared to the state-average. They were mostly located in state's centre and western Sydney. Thirty LGAs with a lower adjusted FTR risk (10% to 30%), accommodating 8% of patients at risk, were mostly found in the southern parts of NSW and Sydney east and south. There were significant spatiotemporal variations of FTR rates across NSW over an eight-year span. Areas identified with significantly high and low FTR risks provide potential opportunities for policy-makers, clinicians and researchers to learn from the success or failure of adopting the best care for surgical patients and build a self-learning organisation and health system.

The choice of spatial scales and spatial smoothness priors for various spatial patterns

Given the drawbacks for using geo-political areas in mapping outcomes unrelated to geo-politics, a compromise is to aggregate and analyse data at the grid level. This has the advantage of allowing spatial smoothing and modelling at a biologically or physically relevant scale. This article addresses two consequent issues: the choice of the spatial smoothness prior and the scale of the grid. Firstly, we describe several spatial smoothness priors applicable for grid data and discuss the contexts in which these priors can be employed based on different aims. Two such aims are considered, i.e., to identify regions with clustering and to model spatial dependence in the data. Secondly, the choice of the grid size is shown to depend largely on the spatial patterns. We present a guide on the selection of spatial scales and smoothness priors for various point patterns based on the two aims for spatial smoothing.

Small area associations between social context and alcohol-attributable mortality in a middle income country

BACKGROUND

Little is known about the association between alcohol-attributable mortality and small area socioeconomic variables when considering causes both wholly and partially attributable to alcohol.

METHODS

An ecological study was conducted of the entire Chilean population aged 15 and older in 345 municipalities nationwide between 2004 and 2009. Deaths were attributed to alcohol consumption either wholly or partially, along with the estimated attributable fractions for each specified cause. Each municipality was characterized according to its average income and educational attainment. Estimates of the ecological associations were produced using a hierarchical Bayesian model, separating out deaths caused by alcohol and dividing them into seven groups of causes.

RESULTS

Alcohol-attributable mortality risk showed an inverse association with income and education at the ecological level. A one-quintile increase in income was associated with an average decrease in risk of 10% (CI 95%: 10-20%) for cardiovascular deaths, 8% (6-10%) for intentional injuries and 7% (3-11%) for unintentional injuries. No associations were found between deaths due to cancers and other causes with income and education.

CONCLUSIONS

Municipalities with lower income and education have higher risk of alcohol-attributable mortality in Chile.

Space-time stick-breaking processes for small area disease cluster estimation

We propose a space-time stick-breaking process for the disease cluster estimation. The dependencies for spatial and temporal effects are introduced by using space-time covariate dependent kernel stick-breaking processes. We compared this model with the space-time standard random effect model by checking each model's ability in terms of cluster detection of various shapes and sizes. This comparison was made for simulated data where the true risks were known. For the simulated data, we have observed that space-time stick-breaking process performs better in detecting medium- and high-risk clusters. For the real data, county specific low birth weight incidences for the state of South Carolina for the years 1997-2007, we have illustrated how the proposed model can be used to find grouping of counties of higher incidence rate.

Use of Poisson spatiotemporal regression models for the Brazilian Amazon Forest: malaria count data

INTRODUCTION

Malaria is a serious problem in the Brazilian Amazon region, and the detection of possible risk factors could be of great interest for public health authorities. The objective of this article was to investigate the association between environmental variables and the yearly registers of malaria in the Amazon region using bayesian spatiotemporal methods.

METHODS

We used Poisson spatiotemporal regression models to analyze the Brazilian Amazon forest malaria count for the period from 1999 to 2008. In this study, we included some covariates that could be important in the yearly prediction of malaria, such as deforestation rate. We obtained the inferences using a bayesian approach and Markov Chain Monte Carlo (MCMC) methods to simulate samples for the joint posterior distribution of interest. The discrimination of different models was also discussed.

RESULTS

The model proposed here suggests that deforestation rate, the number of inhabitants per km², and the human development index (HDI) are important in the prediction of malaria cases.

CONCLUSIONS

It is possible to conclude that human development, population growth, deforestation, and their associated ecological alterations are conducive to increasing malaria risk. We conclude that the use of Poisson regression models that capture the spatial and temporal effects under the bayesian paradigm is a good strategy for modeling malaria counts.