Analysis of Pollution Hazard Intensity: A Spatial Epidemiology Case Study of Soil Pb Contamination.

Heavy industrialization has resulted in the contamination of soil by metals from anthropogenic sources in Anniston, Alabama. This situation calls for increased public awareness of the soil contamination issue and better knowledge of the main factors contributing to the potential sources contaminating residential soil. The purpose of this spatial epidemiology research is to describe the effects of physical factors on the concentration of lead (Pb) in soil in Anniston AL, and to determine the socioeconomic and demographic characteristics of those residing in areas with higher soil contamination. Spatial regression models are used to account for spatial dependencies using these explanatory variables. After accounting for covariates and multicollinearity, results of the analysis indicate that lead concentration in soils varies markedly in the vicinity of a specific foundry (Foundry A), and that proximity to railroads explained a significant amount of spatial variation in soil lead concentration. Moreover, elevated soil lead levels were identified as a concern in industrial sites, neighborhoods with a high density of old housing, a high percentage of African American population, and a low percent of occupied housing units. The use of spatial modelling allows for better identification of significant factors that are correlated with soil lead concentrations.

Analysis of heavy metal sources in soil using kriging interpolation on principal components.

Anniston, Alabama has a long history of operation of foundries and other heavy industry. We assessed the extent of heavy metal contamination in soils by determining the concentrations of 11 heavy metals (Pb, As, Cd, Cr, Co, Cu, Mn, Hg, Ni, V, and Zn) based on 2046 soil samples collected from 595 industrial and residential sites. Principal Component Analysis (PCA) was adopted to characterize the distribution of heavy metals in soil in this region. In addition, a geostatistical technique (kriging) was used to create regional distribution maps for the interpolation of nonpoint sources of heavy metal contamination using geographical information system (GIS) techniques. There were significant differences found between sampling zones in the concentrations of heavy metals, with the exception of the levels of Ni. Three main components explaining the heavy metal variability in soils were identified. The results suggest that Pb, Cd, Cu, and Zn were associated with anthropogenic activities, such as the operations of some foundries and major railroads, which released these heavy metals, whereas the presence of Co, Mn, and V were controlled by natural sources, such as soil texture, pedogenesis, and soil hydrology. In general terms, the soil levels of heavy metals analyzed in this study were higher than those reported in previous studies in other industrial and residential communities.

Weight estimation and significance testing for three focused statistics.

Focused tests for clustering are designed to determine whether there is statistical evidence for raised incidence of some phenomenon around a prespecified location. The tests require definition of what is meant by 'around' the location, and this is achieved by specifying weights associated with surrounding locations. Different weight specifications will yield different levels of statistical significance, and because of the difficulty in knowing how to define the weights, it is tempting to try different definitions with the hope of finding one that is highly significant. This, however, introduces the problem of multiple testing; one will eventually be able to reject a null hypothesis if one tries often enough. This article describes approaches for adjusting the significance level when multiple tests, associated with varying definitions for the weights, are carried out. The approaches are developed for a local scan statistic, a maximum chi-square statistic, and a modified version of Stone's statistic. An illustration is provided using leukemia data from central New York State.

Clusters in irregular areas and lattices.

Geographic areas of different sizes and shapes of polygons that represent counts or rate data are often encountered in social, economic, health, and other information. Often political or census boundaries are used to define these areas because the information is available only for those geographies. Therefore, these types of boundaries are frequently used to define neighborhoods in spatial analyses using geographic information systems and related approaches such as multilevel models. When point data can be geocoded, it is possible to examine the impact of polygon shape on spatial statistical properties, such as clustering. We utilized point data (alcohol outlets) to examine the issue of polygon shape and size on visualization and statistical properties. The point data were allocated to regular lattices (hexagons and squares) and census areas for zip-code tabulation areas and tracts. The number of units in the lattices was set to be similar to the number of tract and zip-code areas. A spatial clustering statistic and visualization were used to assess the impact of polygon shape for zip- and tract-sized units. Results showed substantial similarities and notable differences across shape and size. The specific circumstances of a spatial analysis that aggregates points to polygons will determine the size and shape of the areal units to be used. The irregular polygons of census units may reflect underlying characteristics that could be missed by large regular lattices. Future research to examine the potential for using a combination of irregular polygons and regular lattices would be useful.

Optimal geographic scales for local spatial statistics.

Local spatial statistics are used to test for spatial association in some variable of interest, and to test for clustering around predefined locations. Such statistics require that a neighbourhood be defined around the location of interest. This is done by specifying weights for surrounding regions, and this is tantamount to specification of the scale at which the local dependence or clustering is tested. In practice, weights are usually assigned exogenously, with little thought given to their definition. Most common is the definition of binary adjacency - weights are set equal to one if the region is adjacent to the focal region and to zero otherwise. But this implies a spatial scale that may or may not be the best one to evaluate the variable under study - the actual scale of dependence or clustering is one that is smaller or larger. An alternative strategy is to try different sets of weights corresponding to different spatial scales. The purpose of this article is to provide statistical tests that allow for examination of several local statistics across multiple spatial scales, and yet avoid the need for simulation. Application of these tests leads to a choice of spatial scale through the weights, as well as an assessment of statistical significance. The approach is illustrated using data on leukemia from central New York State.

Optimization of aeromedical base locations in New Mexico using a model that considers crash nodes and paths.

In a recent paper, Tokar Erdemir et al. (2008) introduce models for service systems with service requests originating from both nodes and paths. We demonstrate how to apply and extend their approach to an aeromedical base location application, with specific focus on the state of New Mexico (NM). The current aeromedical base locations of NM are selected without considering motor vehicle crash paths. Crash paths are the roads on which crashes occur, where each road segment has a weight signifying relative crash occurrence. We analyze the loss in accident coverage and location error for current aeromedical base locations. We also provide insights on the relevance of considering crash paths when selecting aeromedical base locations. Additionally, we look briefly at some of the tradeoff issues in locating additional trauma centers vs. additional aeromedical bases in the current aeromedical system of NM. Not surprisingly, tradeoff analysis shows that by locating additional aeromedical bases, we always attain the required coverage level with a lower cost than with locating additional trauma centers.

Case control study of the geographic variability of exposure to disinfectant byproducts and risk for rectal cancer.

BACKGROUND: Levels of byproducts that result from the disinfection of drinking water vary within a water distribution system. This prompted us to question whether the risk for rectal cancer also varies, depending upon one's long term geographic location within the system. Such a geographic distribution in rectal cancer risk would follow naturally from an association between level of byproduct and rectal cancer risk. We assess the effects of estimated geographic variability in exposure to some of the components of the trihalomethane group of disinfectant byproducts (DBPs) on the odds ratios and probabilities for rectal cancer in white males in a case control study of 128 cases and 253 controls, conducted in Monroe County, Western New York State, U.S.A. The study was designed around health data initially collected at the University at Buffalo (Department of Social and Preventative Medicine) as part of the Upstate New York Diet Study, and trihalomethane (THM) data collected from a separate independent study of THMs conducted by Monroe County Department of Health. Case participants were chosen from hospital pathology records. The controls are disease-free white males between 35-90 years old, living in Monroe County, and chosen from control groups for studies from cancer of five other (unrelated) sites. Using a combination of case control methodology and spatial analysis, the spatial patterns of THMs and individual measures of tap water consumption provide estimates of the effects of ingestion of specific amounts of some DBPs on rectal cancer risk. Trihalomethane (THM) data were used to spatially interpolate levels at the taps of cases and controls, and odds ratios were estimated using logistic regression to assess the effects of estimated THM exposure dose on cancer risk, adjusting for alcohol, dietary beta carotene intake, tap water intake, and total caloric intake. RESULTS: Trihalomethane levels varied spatially within the county; although risk for rectal cancer did not increase with total level of trihalomethanes, increasing levels of the component bromoform (measured in ug/day) did correspond with an increase in odds ratios (OR = 1.85; 95% CI = 1.25 - 2.74) for rectal cancer. The highest quartiles of estimated consumption of bromoform (1.69-15.43 ug/day) led to increased risk for rectal cancer (OR = 2.32; 95% CI = 1.22-4.39). Two other THMs were marginally associated with an increase in risk - chlorodibromomethane (OR = 1.78, 95% CI = 1.00-3.19) and bromodichloromethane (OR = 1.15; 95% CI = 1.00-1.32). CONCLUSION: Levels of THMs in the water distribution system exhibited spatial variation that was partially due to variation in water age. We also observed a geographic pattern of increased risk of rectal cancer in areas with the highest levels of bromoform in the county.

Case-control study of the effects of trihalomethanes on urinary bladder cancer risk.

In this research, the authors examined the relation between the estimated concentrations in drinking water of disinfectant byproduct (DBP) trihalomethanes (THMs) and the risk for urinary bladder cancer in a case-control study of 567 white men aged 35 to 90 years, in western New York State. They used logistic regression to estimate odds ratios (ORS) and to assess the effects of THM consumption on cancer risk. Higher levels of consumption of THMs led to increased risk for cancer of the urinary bladder (total 551, a composite measure of THMs based upon method 551 developed by the US Environmental Protection Agency: OR = 2.34; 95% confidence interval [CI] = 1.01-3.66). Results were most significant for bromoform (OR = 3.05; 95% CI = 1.51-5.69), and risk was highest (OR = 5.85; 95% CI = 1.93-17.46) for those who consumed the greatest amount of water at points within the distribution system with the oldest postdisinfected tap water.

Statistical methods for the detection of spatial clustering in case-control data.

In this paper, I develop new approaches for the detection of spatial clustering in case-control data. One method is based upon drawing Thiessen polygons around each control. It is unnecessary to actually draw or compute the boundaries of the polygons; it is sufficient to count, for each control, the number of cases that are closer to that control than to any other control. A second method is similar to the Cuzick-Edwards method, which is based on counts of cases that are among the k-nearest neighbours of cases, but is instead based upon the number of cases within a specified distance of cases. These first two methods are global methods in the sense that they provide a single statistic that measures the degree of spatial clustering. The third method suggests a local statistic, for tests of the null hypothesis of no spatial clustering around a prespecified focus. The method is based upon the cumulative chi2 test, which is typically used to test whether cases are more prevalent than expected around a prespecified location. This is also extended to the case where all observational locations are considered as potential cluster locations and multiple testing is carried out. Each of the new methods is illustrated using data on childhood leukaemia and lymphoma cases in North Humberside.

Recent changes in the spatial pattern of prostate cancer in the U.S.

INTRODUCTION: Spatial-temporal trends in prostate cancer mortality are of interest because of the introduction and increasing use of the prostate-specific antigen (PSA) screening test after 1986. This article describes spatial-temporal changes in U.S. prostate cancer mortality from 1968 to 1998. METHODS: Prostate cancer mortality data were obtained from Compressed Mortality Files available from the National Center for Health Statistics. To minimize potential problems such as small numbers or missing data, the analysis was limited to white males aged 25 and over, and located in 2970 counties with complete data. Statistical analyses included the global distance between observed and expected multinomial probabilities, Hoover's Index of Concentration, and a retrospective test for change in spatial patterns. RESULTS: Fairly steady declines were observed in prostate cancer mortality from 1968 until 1993, with an increasing tendency toward spatial uniformity. Spatial concentration increased from 1994 to 1998, and by 1998 the level of spatial concentration had returned to levels that prevailed during the early to mid-1980s. Comparing 1991-1998 to 1968-1990, the observed number of prostate deaths increased the most rapidly with respect to the expected number in western Appalachia and the south central U.S. Recent relative declines in mortality were observed in southern California and parts of Florida. CONCLUSIONS: The observed results are generally consistent with prior evaluations of prostate cancer spatial-temporal patterns. However, the current study identified a heretofore unnoticed recent pattern of change in western Appalachia and the south central U.S. Recent declines in Florida and southern California may have contributed to recent increases in spatial concentration of prostate cancer mortality, and may possibly be associated with realized benefits from screening programs.

Geographical variation of cerebrovascular disease in New York State: the correlation with income.

BACKGROUND: Income is known to be associated with cerebrovascular disease; however, little is known about the more detailed relationship between cerebrovascular disease and income. We examined the hypothesis that the geographical distribution of cerebrovascular disease in New York State may be predicted by a nonlinear model using income as a surrogate socioeconomic risk factor. RESULTS: We used spatial clustering methods to identify areas with high and low prevalence of cerebrovascular disease at the ZIP code level after smoothing rates and correcting for edge effects; geographic locations of high and low clusters of cerebrovascular disease in New York State were identified with and without income adjustment. To examine effects of income, we calculated the excess number of cases using a non-linear regression with cerebrovascular disease rates taken as the dependent variable and income and income squared taken as independent variables. The resulting regression equation was: excess rate = 32.075-1.22 x 10(-4)(income)+ 8.068x10(-10)(income2), and both income and income squared variables were significant at the 0.01 level. When income was included as a covariate in the non-linear regression, the number and size of clusters of high cerebrovascular disease prevalence decreased. Some 87 ZIP codes exceeded the critical value of the local statistic yielding a relative risk of 1.2. The majority of low cerebrovascular disease prevalence geographic clusters disappeared when the non-linear income effect was included. For linear regression, the excess rate of cerebrovascular disease falls with income; each 10,000 dollars increase in median income of each ZIP code resulted in an average reduction of 3.83 observed cases. The significant nonlinear effect indicates a lessening of this income effect with increasing income. CONCLUSION: Income is a non-linear predictor of excess cerebrovascular disease rates, with both low and high observed cerebrovascular disease rate areas associated with higher income. Income alone explains a significant amount of the geographical variance in cerebrovascular disease across New York State since both high and low clusters of cerebrovascular disease dissipate or disappear with income adjustment. Geographical modeling, including non-linear effects of income, may allow for better identification of other non-traditional risk factors.

Population distribution and redistribution of the baby-boom cohort in the United States: recent trends and implications.

Over 70 million people were born into the baby-boom cohort between 1946 and 1964. Over 65 million of these individuals are presently alive, and thus the cohort continues to exert a powerful influence on regional population change in the United States. In this article, we examine the recent and current geographic distribution of the baby-boom cohort. In 1990, the members of the cohort comprised a particularly high proportion of the population in a small number of dynamic metropolitan areas. We also highlight the recent migration trends exhibited by this cohort; these trends are potentially important early indicators of the retirement-related migration patterns that the cohort might follow. The spatial redistribution of the cohort has many implications, including potentially significant consequences for intergenerational relationships and caregiving. Also highlighted in the article are the temporal and geographical implications for intergenerational caregiving. There has been much attention given to the "sandwich" generation, with its members having dual caregiving responsibilities to both parents and children. A more appropriate designation may be the "stretched" generation, because caregiving seems to extend over a long period. In particular, many members of the baby-boom cohort are beginning to care for their aging parents just as they finish child rearing.

Use of CUSUM and Shewhart charts to monitor regional trends of birth defect reports in New York State.

BACKGROUND: Cumulative sum (CUSUM) charts were originally developed for industrial quality control, but may be adapted for the surveillance of health outcome data, such as birth defects. The reported prevalence of birth defects can vary due to differences in case ascertainment, surveillance practices, or true changes in prevalence. We examined the utility of CUSUM and Shewhart charts for detect-ing changes in prevalence of two different birth defect groups. We chose obstructive renal defects because we expected an increase in reporting due to improved diagnosis. We chose oral clefts for comparison because we expected reporting to be unaffected by changes in diagnostic technologies. METHODS: Data from the New York State Congenital Malformations Registry from 1992-1999 were analyzed using self-starting binomial CUSUM and Shewhart charts for four regions of New York State. RESULTS: CUSUM charts show that reports of obstructive urinary defects have increased from 1992-1999 in all regions of New York State. Reports of oral clefts increased only on Long Island. CONCLUSIONS: The CUSUM method proved useful for identifying changes in birth defect reporting and was able to detect the expected increases in obstructive renal defects. The apparent increase is likely due to improvements in diagnostic imaging techniques. In contrast, we only detected an increase in oral clefts on Long Island, which may be related to under report-ing of cases in the earlier years. CUSUM charts are useful in detecting small, sustained increases in prevalences over time while Shewhart charts are easier to interpret and can detect large sharp increases.

Assessing spatio-temporal variability of risk surfaces using residential history data in a case control study of breast cancer.

BACKGROUND: Most analyses of spatial clustering of disease have been based on either residence at the time of diagnosis or current residence. An underlying assumption in these analyses is that residence can be used as a proxy for environmental exposure. However, exposures earlier in life and not just those in the most recent period may be of significance. In breast cancer, there is accumulating evidence that early life exposures may contribute to risk. We explored spatio-temporal patterns of risk surfaces using data on lifetime residential history in a case control study of breast cancer, and identified elevated areas of risk and areas potentially having more exposure opportunities, defined as risk surfaces in this study. This approach may be more relevant in understanding the environmental etiology of breast cancer, since lifetime cumulative exposures or exposures at critical times may be more strongly associated with risk for breast cancer than exposures from the recent period. RESULTS: A GIS-based exploratory spatial analysis was applied, and spatio-temporal variability of those risk surfaces was evaluated using the standardized difference in density surfaces between cases and controls. The significance of the resulting risk surfaces was tested and reported as p-values. These surfaces were compared for premenopausal and postmenopausal women, and were obtained for each decade, from the 1940s to 1990s. We found strong evidence of clustering of lifetime residence for premenopausal women (for cases relative to controls), and a less strong suggestion of such clustering for postmenopausal women, and identified a substantial degree of temporal variability of the risk surfaces. CONCLUSION: We were able to pinpoint geographic areas with higher risk through exploratory spatial analyses, and to assess temporal variability of the risk surfaces, thus providing a working hypothesis on breast cancer and environmental exposures. Geographic areas with higher case densities need further epidemiologic investigation for potential relationships between lifetime environmental exposures and breast cancer risk. Examination of lifetime residential history provided additional information on geographic areas associated with higher risk; limiting exploration of chronic disease clustering to current residence may neglect important relationships between location and disease.

Geographic clustering of residence in early life and subsequent risk of breast cancer (United States).

OBJECTIVE: This study focused on geographic clustering of breast cancer based on residence in early life and identified spatio-temporal clustering of cases and controls. METHODS: Data were drawn from the WEB study (Western New York Exposures and Breast Cancer Study), a population-based case-control study of incident, pathologically confirmed breast cancer (1996-2001) in Erie and Niagara counties. Controls were frequency-matched to cases on age, race, and county of residence. All cases and controls used in the study provided lifetime residential histories. The k-function difference between cases and controls was used to identify spatial clustering patterns of residence in early life. RESULTS: We found that the evidence for clustered residences at birth and at menarche was stronger than that for first birth or other time periods in adult life. Residences for pre-menopausal cases were more clustered than for controls at the time of birth and menarche. We also identified the size and geographic location of birth and menarche clusters in the study area, and found increased breast cancer risk for pre-menopausal women whose residence was within the cluster compared to those living elsewhere at the time of birth. CONCLUSION: This study provides evidence that early environmental exposures may be related to breast cancer risk, especially for pre-menopausal women.

Monitoring change in spatial patterns of disease: comparing univariate and multivariate cumulative sum approaches.

Prospective disease surveillance has gained increasing attention, particularly in light of recent concern for quick detection of bioterrorist events. Monitoring of health events has the potential for the detection of such events, but the benefits of surveillance extend much more broadly to the quick detection of change in public health. In this paper, univariate and multivariate cumulative sum methods for disease surveillance are compared. Although the univariate method has been previously used in the context of health surveillance, the multivariate method has not. The univariate approach consists of simultaneously and independently monitoring the disease rate in each region; the multivariate approach accounts explicitly for any covariation between regions. The univariate approaches are limited by their lack of ability to account for the spatial autocorrelation of regional data; the multivariate methods are limited by the difficulty in accurately specifying the multiregional covariance structure. The methods are illustrated using both simulated data and county-level data on breast cancer in the northeastern United States. When the degree of spatial autocorrelation is low, the univariate method is generally better at detecting changes in rates that occur in a small number of regions; the multivariate is better when change occurs in a large number of regions.

Approaches to syndromic surveillance when data consist of small regional counts.

INTRODUCTION: Statistical systems designed for syndromic surveillance often must be able to monitor data received simultaneously from multiple regions. Such data might be of limited size, which would eliminate the possibility of using more common surveillance methods that assume data from a normal distribution. OBJECTIVES: The objectives of this study were to design and illustrate a multiregional surveillance system based on data inputs consisting of small regional counts, where frequencies are typically on the order of

Evaluating the reliability of automated collision notification systems.

The use of an automated collision notification (ACN) device in vehicles can greatly reduce the time between crash occurrence and notification of emergency medical services (EMSs). Most ACN devices rely on cellular technology to report important crash information to the proper authorities. The objective of this study was to examine the ability of the existing western New York cellular analog system to support ACN systems. The first task was to develop a model predicting the probability of successfully completing an emergency ACN call at attenuated levels of received signal strength indicator (RSSI), a measurement of the bond between cell phone and tower. Then, empirical estimates were made of the time necessary for call completion at given levels of the RSSI. The RSSI is sampled at locations throughout Erie County, New York, and this information is used to determine the probability of successful call completion for different locations within the county. This model was then applied to historic data for selected past crashes. Finally, the findings were compared with real-world crash data obtained from the ACN Field Operational Test program, where 750 ACN devices were installed in cars and their performance examined over time. An interpolated map of the sampled RSSI values suggests that cellular coverage in Erie County is adequate to support the automated collision network technology. The models and techniques described here are applicable to other areas and regions of the country.

The effects of migration on the detection of geographic differences in disease risk.

Human migration can make it more difficult to detect geographic differences in disease risk because of the spatial diffusion of people originally exposed in a given geographic area. There are also situations where migration can facilitate the detection of disease attributable to environmental hazards. This paper assesses the effects that migration has on the ability to detect regional variability in disease risk. Several characteristics of migration are discussed, including some that are not widely known. Because of regional variations in mobility rates and other characteristics of the migration process, there is substantial regional variation in the ability to detect spatial variation in risk.