Andrew B. Lawson

Disease mapping with WinBUGS and MLwiN

Preface. Notation. 0.1 Standard notation for multilevel modelling. 0.2 Spatial multiple-membership models and the MMMC notation. 0.3 Standard notation for WinBUGS models. 1. Disease mapping basics. 1.1 Disease mapping and map reconstruction. 1.2 Disease map restoration. 2. Bayesian hierarchical modelling. 2.1 Likelihood and posterior distributions. 2.2 Hierarchical models. 2.3 Posterior inference. 2.4 Markov chain Monte Carlo methods. 2.5 Metropolis and Metropolis-Hastings algorithms. 2.6 Residuals and goodness of fit. 3. Multilevel modelling. 3.1 Continuous response models. 3.2 Estimation procedures for multilevel models. 3.3 Poisson response models. 3.4 Incorporating spatial information. 3.5 Discussion. 4. WinBUGS basics. 4.1 About WinBUGS. 4.2 Start using WinBUGS. 4.3 Specification of the model. 4.4 Model fitting. 4.5 Scripts. 4.6 Checking convergence. 4.7 Spatial modelling: GeoBUGS. 4 .8 Conclusions. 5. MLwiN basics. 5.1 About MLwiN. 5.2 Getting started. 5.3 Fitting statistical models. 5.4 MCMC estimation in MLwiN. 5.5 Spatial modelling. 5.6 Conclusions. 6. Relative risk estimation. 6.1 Relative risk estimation using WinBUGS. 6.2 Spatial prediction. 6.3 An analysis of the Ohio dataset using MLwiN. 7. Focused clustering: the analysis of putative health hazards. 7.1 Introduction. 7.2 Study design. 7.3 Problems of inference. 7.4 Modelling the hazard exposure risk. 7.5 Models for count data. 7.6 Bayesian models. 7.7 Focused clustering in WinBUGS. 7.8 Focused clustering in MLwiN. 8. Ecological analysis. 8.1 Introduction. 8.2 Statistical models. 8.3 WinBUGS analyses of ecological datasets. 8.4 MLwiN analyses of ecological datasets. 9. Spatially-correlated survival analysis. 9.1 Survival analysis in WinBUGS. 9.2 Survival analysis in MLwiN. 10. Epilogue. Appendix 1: WinBUGS code for focused clustering models. A.1: Falkirk example. A.2: Ohio example. Appendix 2: S-Plus function for conversion to GeoBUGS format. Bibliography. Index.

Validation of 3 Food Outlet Databases: Completeness and Geospatial Accuracy in Rural and Urban Food Environments

Despite interest in the built food environment, little is known about the validity of commonly used secondary data. The authors conducted a comprehensive field census identifying the locations of all food outlets using a handheld global positioning system in 8 counties in South Carolina (2008–2009). Secondary data were obtained from 2 commercial companies, Dun & Bradstreet, Inc. (D&B) (Short Hills, New Jersey) and InfoUSA, Inc. (Omaha, Nebraska), and the South Carolina Department of Health and Environmental Control (DHEC). Sensitivity, positive predictive value, and geospatial accuracy were compared. The field census identified 2,208 food outlets, significantly more than the DHEC (n = 1,694), InfoUSA (n = 1,657), or D&B (n = 1,573). Sensitivities were moderate for DHEC (68%) and InfoUSA (65%) and fair for D&B (55%). Combining InfoUSA and D&B data would have increased sensitivity to 78%. Positive predictive values were very good for DHEC (89%) and InfoUSA (86%) and good for D&B (78%). Geospatial accuracy varied, depending on the scale: More than 80% of outlets were geocoded to the correct US Census tract, but only 29%–39% were correctly allocated within 100 m. This study suggests that the validity of common data sources used to characterize the food environment is limited. The marked undercount of food outlets and the geospatial inaccuracies observed have the potential to introduce bias into studies evaluating the impact of the built food environment.

Identifying natural and anthropogenic sources of metals in urban and rural soils using GIS-based data, PCA, and spatial interpolation.

Determining sources of neurotoxic metals in rural and urban soils is important for mitigating human exposure. Surface soil from four areas with significant clusters of mental retardation and developmental delay (MR/DD) in children, and one control site were analyzed for nine metals and characterized by soil type, climate, ecological region, land use and industrial facilities using readily available GIS-based data. Kriging, principal component analysis (PCA) and cluster analysis (CA) were used to identify commonalities of metal distribution. Three MR/DD areas (one rural and two urban) had similar soil types and significantly higher soil metal concentrations. PCA and CA results suggested that Ba, Be and Mn were consistently from natural sources; Pb and Hg from anthropogenic sources; and As, Cr, Cu, and Ni from both sources. Arsenic had low commonality estimates, was highly associated with a third PCA factor, and had a complex distribution, complicating mitigation strategies to minimize concentrations and exposures.

Soil metal concentrations and toxicity: associations with distances to industrial facilities and implications for human health.

Urban and rural areas may have different levels of environmental contamination and different potential sources of exposure. Many metals, i.e., arsenic (As), lead (Pb), and mercury (Hg), have well-documented negative neurological effects, and the developing fetus and young children are particularly at risk. Using a database of mother and child pairs, three areas were identified: a rural area with no increased prevalence of mental retardation and developmental delay (MR/DD) (Area A), and a rural area (Area B) and an urban area (Area C) with significantly higher prevalence of MR/DD in children as compared to the state-wide average. Areas were mapped and surface soil samples were collected from nodes of a uniform grid. Samples were analyzed for As, barium (Ba), beryllium (Be), chromium (Cr), copper (Cu), Pb, manganese (Mn), nickel (Ni), and Hg concentrations and for soil toxicity, and correlated to identify potential common sources. ArcGIS was used to determine distances between sample locations and industrial facilities, which were correlated with both metal concentrations and soil toxicity. Results indicated that all metal concentrations (except Be and Hg) in Area C were significantly greater than those in Areas A and B (p< or =0.0001) and that Area C had fewer correlations between metals suggesting more varied sources of metals than in rural areas. Area C also had a large number of facilities whose distances were significantly correlated with metals, particularly Cr (maximum r=0.33; p=0.0002), and with soil toxicity (maximum r=0.25; p=0.007) over a large spatial scale. Arsenic was not associated with distance to any facility and may have a different anthropogenic, or natural source. In contrast to Area C, both rural areas had lower concentrations of metals, lower soil toxicity, and a small number of facilities with significant associations between distance and soil metals.

Metal concentrations in rural topsoil in South Carolina: Potential for human health impact

Rural areas are often considered to have relatively uncontaminated soils; however few studies have measured metals in surface soil from low population areas. Many metals, i.e., arsenic (As), lead (Pb), and mercury (Hg), have well-documented negative neurological effects, and the developing fetus and young children are particularly at risk. Using a Medicaid database, two areas were identified: one with no increased prevalence of mental retardation and developmental delay (MR/DD) (Strip 1) and one with significantly higher prevalence of MR/DD (Strip 2) in children compared to the state-wide average. These areas were mapped and surface soil samples were collected from 0-5 cm depths from nodes of a uniform grid laid out across the sampling areas. Samples were analyzed for As, barium (Ba), beryllium (Be), chromium (Cr), copper (Cu), Pb, manganese (Mn), nickel (Ni), and Hg. Inverse distance weighting (IDW) was used to estimate concentrations throughout each strip area, and a principal component analysis (PCA) was used to identify common sources. All metal concentrations in Strip 2, the MR/DD cluster area, were significantly greater than those in Strip 1 and similar to those found in more urban and highly agricultural areas. Both Strips 1 and 2 had a high number of significant correlations between metals (33 for Strip 1 and 25 for Strip 2), suggesting possible similar natural or anthropogenic sources which was corroborated by PCA. While exposures were not assessed and direct causation between environmental soil metal concentrations and MR/DD cannot be concluded, the high metal concentrations in areas with an elevated prevalence of MR/DD warrants further consideration.

Neighborhood level risk factors for type 1 diabetes in youth: the SEARCH case-control study.

BackgroundEuropean ecologic studies suggest higher socioeconomic status is associated with higher incidence of type 1 diabetes. Using data from a case-control study of diabetes among racially/ethnically diverse youth in the United States (U.S.), we aimed to evaluate the independent impact of neighborhood characteristics on type 1 diabetes risk. Data were available for 507 youth with type 1 diabetes and 208 healthy controls aged 10-22 years recruited in South Carolina and Colorado in 2003-2006. Home addresses were used to identify Census tracts of residence. Neighborhood-level variables were obtained from 2000 U.S. Census. Multivariate generalized linear mixed models were applied.ResultsControlling for individual risk factors (age, gender, race/ethnicity, infant feeding, birth weight, maternal age, number of household residents, parental education, income, state), higher neighborhood household income (p = 0.005), proportion of population in managerial jobs (p = 0.02), with at least high school education (p = 0.005), working outside the county (p = 0.04) and vehicle ownership (p = 0.03) were each independently associated with increased odds of type 1 diabetes. Conversely, higher percent minority population (p = 0.0003), income from social security (p = 0.002), proportion of crowded households (0.0497) and poverty (p = 0.008) were associated with a decreased odds.ConclusionsOur study suggests that neighborhood characteristics related to greater affluence, occupation, and education are associated with higher type 1 diabetes risk. Further research is needed to understand mechanisms underlying the influence of neighborhood context.

Review of methods for space-time disease surveillance.

Abstract A review of some methods for analysis of space–time disease surveillance data is presented. Increasingly, surveillance systems are capturing spatial and temporal data on disease and health outcomes in a variety of public health contexts. A vast and growing suite of methods exists for detection of outbreaks and trends in surveillance data and the selection of appropriate methods in a given surveillance context is not always clear. While most reviews of methods focus on algorithm performance, in practice, a variety of factors determine what methods are appropriate for surveillance. In this review, we focus on the role of contextual factors such as scale, scope, surveillance objective, disease characteristics, and technical issues in relation to commonly used approaches to surveillance. Methods are classified as testing-based or model-based approaches. Reviewing methods in the context of factors other than algorithm performance highlights important aspects of implementing and selecting appropriate disease surveillance methods.

Probability of intellectual disability is associated with soil concentrations of arsenic and lead

BACKGROUND The association between metals in water and soil and adverse child neurologic outcomes has focused on the singular effect of lead (Pb), mercury (Hg), and arsenic (As). This study describes the complex association between soil concentrations of As combined with Pb and the probability of intellectual disability (ID) in children. METHODS We used a retrospective cohort design with 3988 mother child pairs who were insured by Medicaid and lived during pregnancy and early childhood in South Carolina between 1/1/97 and 12/31/02. The children were followed until 6/1/08, using computerized service files, to identify the diagnosis of ID in medical records and verified by either school placement or disability service records. The soil was sampled using a uniform grid and analyzed for eight metals. The metal concentrations were interpolated using Bayesian Kriging to estimate concentration at individual residences. RESULTS The probability of ID increased for increasing concentrations of As and Pb in the soil. The Odds Ratio for ID, for one unit change in As was 1.130 (95% confidence interval 1.048-1.218) for Pb was 1.002 (95% confidence interval 1.000-1.004). We identified effect modification for the infants based on their birth weight for gestational age status and only infants who were normal size for their gestational age had increased probability of ID based on the As and Pb soil concentrations (OR for As at normal weight for gestational age=1.151 (95% CI: 1.061-1.249) and OR for Pb at normal for gestational age=1.002 (95% CI: 1.002-1.004)). For normal weight for gestational age children when As=22 mg kg(-1) and Pb=200 mg kg(-1) the risk for ID was 11% and when As=22 mg kg(-1)and Pb=400 mg kg(-1) the probability of ID was 65%. CONCLUSION The probability of ID is significantly associated with the interaction between Pb and As for normal weight for gestational age infants.

The relationship between mental retardation and developmental delays in children and the levels of arsenic, mercury and lead in soil samples taken near their mother’s residence during pregnancy☆

This study was designed to evaluate the association between lead, mercury, and arsenic in the soil near maternal residences during pregnancy and mental retardation or developmental disability (MR/DD) in children. The study was conducted using 6,048 mothers who did not move throughout their pregnancies and lived within six strips of land in South Carolina and were insured by Medicaid between January 1, 1997 and December 31, 2002. The mother child pairs were then followed until June 1, 2008, through their Medicaid reimbursement files, to identify children diagnosed with MR/DD. The soil was sampled for mercury (Hg), lead (Pb), and As based on a uniform grid, and the soil concentrations were Kriged to estimate chemical concentration at individual locations. We identified a significant relationship between MR/DD and As, and the form of the relationship was nonlinear, after controlling for other known risk factors.

Characterizing the food retail environment: Impact of count, type, and geospatial error in 2 secondary data sources

OBJECTIVE Commercial listings of food retail outlets are increasingly used by community members and food policy councils and in multilevel intervention research to identify areas with limited access to healthier food. This study quantified the amount of count, type, and geospatial error in 2 commercial data sources. METHODS InfoUSA and Dun and Bradstreet were compared with a validated field census and validity statistics were calculated. RESULTS Considering only completeness, Dun and Bradstreet data undercounted 24% of existing supermarkets and grocery stores, and InfoUSA, 29%. In addition, considering accuracy of outlet type assignment increased the undercount error to 42% and 39%, respectively. Marked overcount existed as well, and only 43% of existing supermarkets were correctly identified with respect to presence, outlet type, and location. CONCLUSIONS AND IMPLICATIONS Relying exclusively on secondary data to characterize the food environment will result in substantial error. Whereas extensive data cleaning can offset some error, verification of outlets with a field census is still the method of choice.