Spatial Poisson Regression Models Research Articles

Visible minority status and occupation were associated with increased COVID-19 infection in Greater Vancouver British Columbia between June and November 2020: an ecological study.

The COVID-19 pandemic has highlighted health disparities, especially among specific population groups. This study examines the spatial relationship between the proportion of visible minorities (VM), occupation types and COVID-19 infection in the Greater Vancouver region of British Columbia, Canada. Provincial COVID-19 case data between June 24, 2020, and November 7, 2020, were aggregated by census dissemination area and linked with sociodemographic data from the Canadian 2016 census. Bayesian spatial Poisson regression models were used to examine the association between proportion of visible minorities, occupation types and COVID-19 infection. Models were adjusted for COVID-19 testing rates and other sociodemographic factors. Relative risk (RR) and 95% Credible Intervals (95% CrI) were calculated. We found an inverse relationship between the proportion of the Chinese population and risk of COVID-19 infection (RR = 0.98 95% CrI = 0.96, 0.99), whereas an increased risk was observed for the proportions of the South Asian group (RR = 1.10, 95% CrI = 1.08, 1.12), and Other Visible Minority group (RR = 1.06, 95% CrI = 1.04, 1.08). Similarly, a higher proportion of frontline workers (RR = 1.05, 95% CrI = 1.04, 1.07) was associated with higher infection risk compared to non-frontline. Despite adjustments for testing, housing, occupation, and other social economic status variables, there is still a substantial association between the proportion of visible minorities, occupation types, and the risk of acquiring COVID-19 infection in British Columbia. This ecological analysis highlights the existing disparities in the burden of diseases among different visible minority populations and occupation types.

Organized abandonment under racial capitalism: Measuring accountable actors of structural racism for public health research and action

Understanding the shifting nature of structural racism historically and across institutions is vital for effective action towards racial health equity. While public health research on structural racism is rapidly increasing, most studies are missing the interdependence of policies and institutional practices over time that shape power imbalances and lead to entrenched health inequities. Here, we discuss Ruth Wilson Gilmore's concept of organized abandonment — the intentional disinvestment in communities which, in turn, creates opportunities for extraction, revenue generation, and carceral enforcement to fill the cracks of a compromised social infrastructure — to encourage action-oriented public health research that is grounded in history and an understanding of racial capitalism. We present a case example using publicly-available data on redlining, gentrification and policing in Seattle, Washington. We mapped the intersections of redlining and gentrification and estimated their neighborhood-level association with police activity using Bayesian spatial Poisson regression models. We found that histories of racist housing policies like redlining and processes of gentrification are interdependent and shape contemporary neighborhood racial and economic segregation and police activity. Compared to structurally advantaged neighborhoods, police stops were higher in neighborhoods that were 1) historically disinvested (i.e. redlined) and remain low-income and structurally disadvantaged and 2) formerly industrial and business districts that were not redlined and are now gentrified. Notably, we found that policing practices were significantly more intensive in neighborhoods that were both high redlined and gentrified. Together, these findings illustrate how the place-based racialized processes of dispossession, displacement and policing are deeply intertwined to maintain racial capitalism. Our findings also highlight the importance of examining multiple racialized processes simultaneously to fill critical gaps in the existing literature that are necessary for sustainable solutions to address structural racism.

On Study of the Occurrence of Earth-Size Planets in Kepler Mission Using Spatial Poisson Model

The problem of determining the occurrence rate for Earth-size planets orbiting Sun-like stars is emerging in the universe. We propose a methodology based on a spatial Poisson regression model with model parameters being inferred by the Bayesian framework to investigate this occurrence rate. We analyzed an exoplanet sample and its corresponding survey completeness data. Our results suggest that 46% of Sun-like stars have an Earth-size (i.e., 1–2 times Earth radii) planet with an orbital period of 5–100 days. Furthermore, we are also interested in the occurrence rate of Earth analogs hosted by GK dwarf stars (i.e., orbital period of 200–400 days and size 1–2 times Earth radii). After completeness correction, we obtained an occurrence rate of 0.18% based on the proposed methodology.

Exploring the geographical distribution of human cryptosporidiosis in Southern Ontario from 2011 to 2014.

Cryptosporidium is a protozoan parasite of increasing global public health concern because of its ability to cause disease in both humans and animals through contaminated food and water supplies. In Canada, most human cryptosporidiosis cases are due to Cryptosporidium hominis; however, the presence of zoonotic Cryptosporidium parvum has been observed. Since 2005, the incidence of cryptosporidiosis in Ontario has been consistently higher than the national average; however, it is not understood why, suggesting an incomplete understanding of the pathogen's ecology, epidemiology and transmission pathways. The goal of this study was to explore the spatial distribution of human cryptosporidiosis across the 29 Public Health Unit (PHU) areas of Southern Ontario from 2011 to 2014. Surveillance data on human cryptosporidiosis were obtained from Public Health Ontario. Choropleth and isopleth maps were used to display the distribution of incidence rates of human cryptosporidiosis. High-rate clusters of human cryptosporidiosis were identified. Poisson and spatial Poisson regression models were used to determine the relationship between the incidence of human cryptosporidiosis, cattle density and the smoothed farm-level prevalence of bovine cryptosporidiosis at the PHU level. The annual incidence of reported human cryptosporidiosis in Southern Ontario ranged from 1.62 (95% CI: 1.41-1.86) to 1.82 (95%CI: 1.60-2.06) cases per 100,000 population, with an overall cumulative incidence of 6.91 (95%CI: 6.47-7.39) cases per 100,000 for the 4-year study period. High-risk clusters of human cryptosporidiosis were identified in each year. The relative risk for the clusters ranged from 2.03 (95% CI: 1.63-2.55) to 6.87 (95% CI: 5.07-9.30). A relationship was found between the incidence of cryptosporidiosis and dairy cattle density. Based on this study, the Central West region would be an ideal ecological system to conduct further targeted surveillance to identify factors that may be contributing to the higher burden of cryptosporidiosis in the human and bovine populations in the region.

Social determinants and spatio-temporal variation of Ischemic Heart Disease in Manitoba

IntroductionThe aim was to study any spatial and/or temporal patterns of ischemic heart disease (IHD) prevalence and measure the effects of selected social determinants on these spatial and space-time patterns.MethodsData were obtained from the Population Research Data Repository housed at the Manitoba Centre for Health Policy to identify persons who were diagnosed with IHD between 1995 and 2018. These persons were geocoded to 96 geographic regions of Manitoba. An area-level socioeconomic factor index (SEFI-2) and the proportion of the population who was Indigenous were calculated for each geographic region using the 2016 Canadian Census data. Associations between these factors and IHD prevalence were measured using Bayesian spatial Poisson regression models. Temporal trends and spatio-temporal trends were measured using Bayesian spatio-temporal Poisson regression models.ResultsUnivariable models showed a significant association with increased regional Indigenous population proportion associated with a higher prevalence of IHD (RR: 0.07, 95% CredInt: (0.05, 0.10)) and for SEFI-2 (RR: 0.17, 95% CredInt: (0.11, 0.23)). Using a multivariable model, after accounting for the proportion of the population that was Indigenous, there was no evidence of an association between IHD prevalence and area-level socioeconomic factor. Spatio-temporal models showed no significant overall temporal trend in IHD prevalence, but there were significant spatially varying temporal trends within the 96 regions.ConclusionsAssociation between Indigenous population proportion and IHD is consistent with previous research. No significant overall temporal trend was measured. However, regions with significantly increasing trends and significantly decreasing trends in IHD prevalence were identified.

Exposure to urban green space may both promote and harm mental health in socially vulnerable neighborhoods: A neighborhood-scale analysis in New York City

BackgroundThere is growing evidence that exposure to green space can impact mental health, but these effects may be context dependent. We hypothesized that associations between residential green space and mental health can be modified by social vulnerability. MethodWe conducted an ecological cross-sectional analysis to evaluate the effects of green space exposure on mental disorder related emergency room (ER) visits in New York City at the level of census tract. To objectively represent green space exposure at the neighborhood scale, we calculated three green space exposure metrics, namely proximity to the nearest park, percentage of green space, and visibility of greenness. Using Bayesian hierarchical spatial Poisson regression models, we evaluated neighborhood social vulnerability as a potential modifier of greenness-mental disorder associations, while accounting for the spatially correlated structures. ResultsWe found significant associations between green space exposure (involving both proximity and visibility) and total ER visits for mental disorders in neighborhoods with high social vulnerability, but no significant associations in neighborhoods with low social vulnerability. We also identified specific neighborhoods with particularly high ER utilization for mental disorders. ConclusionsOur findings suggest that exposure to green space is associated with ER visits for mental disorders, but that neighborhood social vulnerability can modify this association. Future research is needed to confirm our finding with longitudinal designs at the level of individuals.

Geospatial analysis of pertussis and its risk factors in southern Ontario from 2005–2016

Introduction: Pertussis, commonly known as whooping cough, is a bacterial respiratory tract infection caused by Bordetella pertussis. Pertussis affects more than 48 million people worldwide annually, most of whom are under the age of 5.
 Hypothesis & Objectives: The hypothesis being investigated is that pertussis incidence, between 2005 and 2016, is not equally distributed across public health units in southern Ontario. We aim to identify disease cluster locations and associate geospatial fluctuations in incidence rates with putative risk factors.
 Materials and Methods: Data was sourced from Public Health Ontario on pertussis incidence in southern Ontario for all ages, specifically for each public health unit’s geographical area. A choropleth map was generated using data smoothed by empirical Bayesian estimation in a spatial analysis context. Following the creation of an incidence map for southern Ontario, the spatial scan test was applied to elucidate the existence of any disease clusters at a public health unit level. Moran’s I was used to determine whether there was evidence of any spatial dependence in pertussis incidence. Finally, putative risk factors were assessed in Poisson regression models and spatial Poisson regression models as potential predictor variables.
 Results and Discussion: The flexible spatial scan test identified three spatial clusters where incidence rates of pertussis were higher than expected. A spatial Poisson regression model was fit that included predictor variables of socioeconomic status and population density. For every 100 people/km2 increase in population density there was a significant 6% increase in pertussis incidence (p=0.03). Interestingly, vaccination rates were not found to be predictive of pertussis incidence nor did the variable improve the model. This epidemiological study identifies where pertussis incidence is clustered and what variables it is associated with, both of which are valuable for public health purposes and as a reference for future research into pertussis.

Future Lyme disease risk in the south-eastern United States based on projected land cover.

Lyme disease is the most significant vector-borne disease in the United States, and its southward advance over several decades has been quantified. Previous research has examined the potential role of climate change on the disease's expansion, but no studies have considered the role of future land cover upon its distribution. This research examines Lyme disease risk in the south-eastern U.S. based on projected land cover developed under four Intergovernmental Panel on Climate Change Scenarios: A1B, A2, B1, and B2. Land cover types and edge indices significantly associated with Lyme disease in Virginia were incorporated into a spatial Poisson regression model to quantify potential land cover suitability for Lyme disease in the south-eastern U.S. under each scenario. Our results indicate an intensification of potential land cover suitability for Lyme disease under the A scenarios and a decrease of potential land cover suitability under the B scenarios. The decrease under the B scenarios is a critical result, indicating that Lyme disease risk can be decreased by making different land cover choices. Additionally, health officials can focus efforts in projected high incidence areas.

Epidemiology of canine heartworm (Dirofilaria immitis) infection in domestic dogs in Ontario, Canada: Geographic distribution, risk factors and effects of climate.

Dirofilaria immitis is the causal agent of heartworm, a mosquito-borne parasite that primarily infects domestic and wild canids. The infection is endemic in parts of Canada, and Ontario has been identified as the province where the majority of heartworm infections occur. Test results for blood samples submitted by veterinary clinics for the years 2007-2016 were used to conduct a spatial risk analysis of heartworm among domestic dogs in Ontario. The geographic extent of the apparent heartworm prevalence was examined through smoothed choropleth maps for all 49 census division regions. Furthermore, the regions were assessed for local clusters in apparent prevalence using the flexible spatial scan statistic. Three clusters were found and located in western, southern and eastern Ontario, respectively. A spatial Poisson regression model for heartworm prevalence among pet dog populations in southern Ontario census divisions was fit to determine the association between human population size, heartworm development units (HDUs), climate moisture index (CMI), precipitation and directions, east or north, with heartworm infection. The model identified the spatial distribution of HDUs and CMI as positively associated with heartworm infection and therefore important predictors of the infection. In contrast, human population size, increasing northern latitude and drier areas were negatively associated with heartworm infection. The east direction and precipitation were not significant.

Spatial Variable Selection and An Application to Virginia Lyme Disease Emergence

ABSTRACT Lyme disease is an infectious disease, that is, caused by a bacterium called Borrelia burgdorferi sensu stricto. In the United States, Lyme disease is one of the most common infectious diseases. The major endemic areas of the disease are New England, Mid-Atlantic, East-North Central, South Atlantic, and West North-Central. Virginia is on the front-line of the disease’s diffusion from the northeast to the south. One of the research objectives for the infectious disease community is to identify environmental and economic variables that are associated with the emergence of Lyme disease. In this article, we use a spatial Poisson regression model to link the spatial disease counts and environmental and economic variables, and develop a spatial variable selection procedure to effectively identify important factors by using an adaptive elastic net penalty. The proposed methods can automatically select important covariates, while adjusting for possible spatial correlations of disease counts. The performance of the proposed method is studied and compared with existing methods via a comprehensive simulation study. We apply the developed variable selection methods to the Virginia Lyme disease data and identify important variables that are new to the literature. Supplementary materials for this article, including a standardized description of the materials available for reproducing the work, are available as an online supplement.

Short-Term Effects of Climatic Variables on Hand, Foot, and Mouth Disease in Mainland China, 2008–2013: A Multilevel Spatial Poisson Regression Model Accounting for Overdispersion

BackgroundHand, Foot, and Mouth Disease (HFMD) is a worldwide infectious disease. In China, many provinces have reported HFMD cases, especially the south and southwest provinces. Many studies have found a strong association between the incidence of HFMD and climatic factors such as temperature, rainfall, and relative humidity. However, few studies have analyzed cluster effects between various geographical units.MethodsThe nonlinear relationships and lag effects between weekly HFMD cases and climatic variables were estimated for the period of 2008–2013 using a polynomial distributed lag model. The extra-Poisson multilevel spatial polynomial model was used to model the exact relationship between weekly HFMD incidence and climatic variables after considering cluster effects, provincial correlated structure of HFMD incidence and overdispersion. The smoothing spline methods were used to detect threshold effects between climatic factors and HFMD incidence.ResultsThe HFMD incidence spatial heterogeneity distributed among provinces, and the scale measurement of overdispersion was 548.077. After controlling for long-term trends, spatial heterogeneity and overdispersion, temperature was highly associated with HFMD incidence. Weekly average temperature and weekly temperature difference approximate inverse “V” shape and “V” shape relationships associated with HFMD incidence. The lag effects for weekly average temperature and weekly temperature difference were 3 weeks and 2 weeks. High spatial correlated HFMD incidence were detected in northern, central and southern province. Temperature can be used to explain most of variation of HFMD incidence in southern and northeastern provinces. After adjustment for temperature, eastern and Northern provinces still had high variation HFMD incidence.ConclusionWe found a relatively strong association between weekly HFMD incidence and weekly average temperature. The association between the HFMD incidence and climatic variables spatial heterogeneity distributed across provinces. Future research should explore the risk factors that cause spatial correlated structure or high variation of HFMD incidence which can be explained by temperature. When analyzing association between HFMD incidence and climatic variables, spatial heterogeneity among provinces should be evaluated. Moreover, the extra-Poisson multilevel model was capable of modeling the association between overdispersion of HFMD incidence and climatic variables.

An Examination of the Demographic and Environmental Variables Correlated with Lyme Disease Emergence in Virginia.

Lyme disease is the United States' most significant vector-borne illness. Virginia, on the southern edge of the disease's currently expanding range, has experienced an increase in Lyme disease both spatially and temporally, with steadily increasing rates over the past decade and disease spread from the northern to the southwestern part of the state. This study used a Geographic Information System and a spatial Poisson regression model to examine correlations between demographic and land cover variables, and human Lyme disease from 2006 to 2010 in Virginia. Analysis indicated that herbaceous land cover is positively correlated with Lyme disease incidence rates. Areas with greater interspersion between herbaceous and forested land were also positively correlated with incidence rates. In addition, income and age were positively correlated with incidence rates. Levels of development, interspersion of herbaceous and developed land, and population density were negatively correlated with incidence rates. Abundance of forest fragments less than 2 hectares in area was not significantly correlated. Our results support some findings of previous studies on ecological variables and Lyme disease in endemic areas, but other results have not been found in previous studies, highlighting the potential contribution of new variables as Lyme disease continues to emerge southward.

Impact of air pollution on physician office visits for common childhood conditions in Ontario, Canada

Methods PM2.5 and temperature measurements were obtained from satellite data for all of Ontario [2]. Physician office visits were stratified into two groups based on the literature: air pollution-sensitive (acute respiratory infections, allergic rhinitis, asthma, bronchiolitis, diabetes, otitis media) and air pollution-insensitive (gastroenteritis, injuries). Claims data were obtained for every month in 2010 from health administrative databases for children 0-14 years of age. Ageand sex-standardized morbidity ratios (SMRs) were calculated by region in Ontario. Spatial Poisson regression models were used to analyze the relationship between PM2.5 and physician office visits, with temperature as a covariate. Results Crude rates of physician office visits are presented in Table 1. As expected, fine particulate was significantly associated with monthly rates of physician office visits for air pollution-sensitive conditions, and not for insensitive conditions. Fitted SMRs for air pollution-sensitive conditions are presented in Figure 1. SMRs for sensitive and insensitive conditions were strongly positively correlated (r = 0.53), and data were spatially autocorrelated. This suggests an underlying spatial process that influences physician office visit rates for common childhood conditions, both for air pollution-sensitive and -insensitive conditions.

Disparities in Spatial Prevalence of Feline Retroviruses due to Data Aggregation: A Case of the Modifiable Areal Unit Problem.

The knowledge of the spatial distribution feline immunodeficiency virus and feline leukemia virus infections, which are untreatable, can inform on their risk factors and high-risk areas to enhance control. However, when spatial analysis involves aggregated spatial data, results may be influenced by the spatial scale of aggregation, an effect known as the modifiable areal unit problem (MAUP). In this study, area level risk factors for both infections in 28,914 cats tested with ELISA were investigated by multivariable spatial Poisson regression models along with MAUP effect on spatial clustering and cluster detection (for postal codes, counties, and states) by Moran's I test and spatial scan test, respectively. The study results indicate that the significance and magnitude of the association of risk factors with both infections varied with aggregation scale. Further more, Moran's I test only identified spatial clustering at postal code and county levels of aggregation. Similarly, the spatial scan test indicated that the number, size, and location of clusters varied over aggregation scales. In conclusion, the association between infection and area was influenced by the choice of spatial scale and indicates the importance of study design and data analysis with respect to specific research questions.

Bayesian discrete latent spatial modeling of crack initiation in orthopaedic hip replacement bone cement

In this paper, we propose a spatial model for the initiation of cracks in the bone cement of hip replacement specimens. The failure of hip replacements can be attributed mainly to damage accumulation, consisting of crack initiation and growth, occurring in the cement mantle that interlocks the hip prosthesis and the femur bone. Since crack initiation is an important factor in determining the lifetime of a replacement, the understanding of the reasons for crack initiation is vital in attempting to prolong the life of the hip replacement. The data consist of crack location coordinates from five laboratory experimental models, together with stress measurements. It is known that stress plays a major role in the initiation of cracks, and it is also known that other unmeasurable factors such as air bubbles (pores) in the cement mantle are also influential. We propose an identity-link spatial Poisson regression model for the counts of cracks in discrete regions of the cement, incorporating both the measured (stress), and through a latent process, any unmeasured factors (possibly pores) that may be influential. All analysis is carried out in a Bayesian framework, allowing for the inclusion of prior information obtained from engineers, and parameter estimation for the model is done via Markov chain Monte Carlo techniques.

Does a Gibbs sampler approach to spatial Poisson regression models outperform a single site MH sampler?

A Gibbs sampler for a Poisson regression model including spatial effects is presented and evaluated. The approach is based on that a Poisson regression model can be transformed into an approximate normal linear model by data augmentation using the introduction of two sequences of latent variables. It is shown how this methodology can be extended to spatial Poisson regression models and details of the resulting Gibbs sampler are given. In particular, the influence of model parameterisation and different update strategies on the mixing of the MCMC chains is discussed. The developed Gibbs samplers are analysed in two simulation studies and applied to model the expected number of claims for policyholders of a German car insurance company. The mixing of the Gibbs samplers depends crucially on the model parameterisation and the update schemes. The best mixing is achieved when collapsed algorithms are used, reasonable low autocorrelations for the spatial effects are obtained in this case. For the regression effects however, autocorrelations are rather high, especially for data with very low heterogeneity. For comparison a single component Metropolis–Hastings algorithms is applied which displays very good mixing for all components. Although the Metropolis–Hastings sampler requires a higher computational effort, it outperforms the Gibbs samplers which would have to be run considerably longer in order to obtain the same precision of the parameters.