Positional error in automated geocoding of residential addresses

Positional error and time-activity patterns in near-highway proximity studies: an exposure misclassification analysis

BACKGROUND

The growing interest in research on the health effects of near-highway air pollutants requires an assessment of potential sources of error in exposure assignment techniques that rely on residential proximity to roadways.

METHODS

We compared the amount of positional error in the geocoding process for three different data sources (parcels, TIGER and StreetMap USA) to a "gold standard" residential geocoding process that used ortho-photos, large multi-building parcel layouts or large multi-unit building floor plans. The potential effect of positional error for each geocoding method was assessed as part of a proximity to highway epidemiological study in the Boston area, using all participants with complete address information (N = 703). Hourly time-activity data for the most recent workday/weekday and non-workday/weekend were collected to examine time spent in five different micro-environments (inside of home, outside of home, school/work, travel on highway, and other). Analysis included examination of whether time-activity patterns were differentially distributed either by proximity to highway or across demographic groups.

RESULTS

Median positional error was significantly higher in street network geocoding (StreetMap USA = 23 m; TIGER = 22 m) than parcel geocoding (8 m). When restricted to multi-building parcels and large multi-unit building parcels, all three geocoding methods had substantial positional error (parcels = 24 m; StreetMap USA = 28 m; TIGER = 37 m). Street network geocoding also differentially introduced greater amounts of positional error in the proximity to highway study in the 0-50 m proximity category. Time spent inside home on workdays/weekdays differed significantly by demographic variables (age, employment status, educational attainment, income and race). Time-activity patterns were also significantly different when stratified by proximity to highway, with those participants residing in the 0-50 m proximity category reporting significantly more time in the school/work micro-environment on workdays/weekdays than all other distance groups.

CONCLUSIONS

These findings indicate the potential for both differential and non-differential exposure misclassification due to geocoding error and time-activity patterns in studies of highway proximity. We also propose a multi-stage manual correction process to minimize positional error. Additional research is needed in other populations and geographic settings.

Impact of geocoding methods on associations between long-term exposure to urban air pollution and lung function

BACKGROUND

Errors in address geocodes may affect estimates of the effects of air pollution on health.

OBJECTIVE

We investigated the impact of four geocoding techniques on the association between urban air pollution estimated with a fine-scale (10 m × 10 m) dispersion model and lung function in adults.

METHODS

We measured forced expiratory volume in 1 sec (FEV1) and forced vital capacity (FVC) in 354 adult residents of Grenoble, France, who were participants in two well-characterized studies, the Epidemiological Study on the Genetics and Environment on Asthma (EGEA) and the European Community Respiratory Health Survey (ECRHS). Home addresses were geocoded using individual building matching as the reference approach and three spatial interpolation approaches. We used a dispersion model to estimate mean PM10 and nitrogen dioxide concentrations at each participant's address during the 12 months preceding their lung function measurements. Associations between exposures and lung function parameters were adjusted for individual confounders and same-day exposure to air pollutants. The geocoding techniques were compared with regard to geographical distances between coordinates, exposure estimates, and associations between the estimated exposures and health effects.

RESULTS

Median distances between coordinates estimated using the building matching and the three interpolation techniques were 26.4, 27.9, and 35.6 m. Compared with exposure estimates based on building matching, PM10 concentrations based on the three interpolation techniques tended to be overestimated. When building matching was used to estimate exposures, a one-interquartile range increase in PM10 (3.0 μg/m3) was associated with a 3.72-point decrease in FVC% predicted (95% CI: -0.56, -6.88) and a 3.86-point decrease in FEV1% predicted (95% CI: -0.14, -3.24). The magnitude of associations decreased when other geocoding approaches were used [e.g., for FVC% predicted -2.81 (95% CI: -0.26, -5.35) using NavTEQ, or 2.08 (95% CI -4.63, 0.47, p = 0.11) using Google Maps].

CONCLUSIONS

Our findings suggest that the choice of geocoding technique may influence estimated health effects when air pollution exposures are estimated using a fine-scale exposure model.

Performance of GPS-devices for environmental exposure assessment

Integration of individual time-location patterns with spatially resolved exposure maps enables a more accurate estimation of personal exposures to environmental pollutants than using estimates at fixed locations. Current global positioning system (GPS) devices can be used to track an individual's location. However, information on GPS-performance in environmental exposure assessment is largely missing. We therefore performed two studies. First, a commute-study, where the commute of 12 individuals was tracked twice, testing GPS-performance for five transport modes and two wearing modes. Second, an urban-tracking study, where one individual was tracked repeatedly through different areas, focused on the effect of building obstruction on GPS-performance. The median error from the true path for walking was 3.7 m, biking 2.9 m, train 4.8 m, bus 4.9 m, and car 3.3 m. Errors were larger in a high-rise commercial area (median error=7.1 m) compared with a low-rise residential area (median error=2.2 m). Thus, GPS-performance largely depends on the transport mode and urban built-up. Although ~85% of all errors were <10 m, almost 1% of the errors were >50 m. Modern GPS-devices are useful tools for environmental exposure assessment, but large GPS-errors might affect estimates of exposures with high spatial variability.

A case-referent study: light at night and breast cancer risk in Georgia

BACKGROUND

Literature has identified detrimental health effects from the indiscriminate use of artificial nighttime light. We examined the co-distribution of light at night (LAN) and breast cancer (BC) incidence in Georgia, with the goal to contribute to the accumulating evidence that exposure to LAN increases risk of BC.

METHODS

Using Georgia Comprehensive Cancer Registry data (2000-2007), we conducted a case-referent study among 34,053 BC cases and 14,458 lung cancer referents. Individuals with lung cancer were used as referents to control for other cancer risk factors that may be associated with elevated LAN, such as air pollution, and since this cancer type was not previously associated with LAN or circadian rhythm disruption. DMSP-OLS Nighttime Light Time Series satellite images (1992-2007) were used to estimate LAN levels; low (0-20 watts per sterradian cm(2)), medium (21-41 watts per sterradian cm(2)), high (>41 watts per sterradian cm(2)). LAN levels were extracted for each year of exposure prior to case/referent diagnosis in ArcGIS.

RESULTS

Odds ratios (OR) and 95% confidence intervals (CI) were estimated using logistic regression models controlling for individual-level year of diagnosis, race, age at diagnosis, tumor grade, stage; and population-level determinants including metropolitan statistical area (MSA) status, births per 1,000 women aged 15-50, percentage of female smokers, MSA population mobility, and percentage of population over 16 in the labor force. We found that overall BC incidence was associated with high LAN exposure (OR = 1.12, 95% CI [1.04, 1.20]). When stratified by race, LAN exposure was associated with increased BC risk among whites (OR = 1.13, 95% CI [1.05, 1.22]), but not among blacks (OR = 1.02, 95% CI [0.82, 1.28]).

CONCLUSIONS

Our results suggest positive associations between LAN and BC incidence, especially among whites. The consistency of our findings with previous studies suggests that there could be fundamental biological links between exposure to artificial LAN and increased BC incidence, although additional research using exposure metrics at the individual level is required to confirm or refute these findings.

Inaccuracy, uncertainty and the space-time permutation scan statistic

The space-time permutation scan statistic (STPSS) is designed to identify hot (and cool) spots of space-time interaction within patterns of spatio-temporal events. While the method has been adopted widely in practice, there has been little consideration of the effect inaccurate and/or incomplete input data may have on its results. Given the pervasiveness of inaccuracy, uncertainty and incompleteness within spatio-temporal datasets and the popularity of the method, this issue warrants further investigation. Here, a series of simulation experiments using both synthetic and real-world data are carried out to better understand how deficiencies in the spatial and temporal accuracy as well as the completeness of the input data may affect results of the STPSS. The findings, while specific to the parameters employed here, reveal a surprising robustness of the method's results in the face of these deficiencies. As expected, the experiments illustrate that greater degradation of input data quality leads to greater variability in the results. Additionally, they show that weaker signals of space-time interaction are those most affected by the introduced deficiencies. However, in stark contrast to previous investigations into the impact of these input data problems on global tests of space-time interaction, this local metric is revealed to be only minimally affected by the degree of inaccuracy and incompleteness introduced in these experiments.

A multi-stage approach to maximizing geocoding success in a large population-based cohort study through automated and interactive processes

To enable spatial analyses within a large, prospective cohort study of nearly 86,000 adults enrolled in a 12-state area in the southeastern United States of America from 2002-2009, a multi-stage geocoding protocol was developed to efficiently maximize the proportion of participants assigned an address level geographic coordinate. Addresses were parsed, cleaned and standardized before applying a combination of automated and interactive geocoding tools. Our full protocol increased the non-Post Office (PO) Box match rate from 74.5% to 97.6%. Overall, we geocoded 99.96% of participant addresses, with only 5.2% at the ZIP code centroid level (2.8% PO Box and 2.3% non-PO Box addresses). One key to reducing the need for interactive geocoding was the use of multiple base maps. Still, addresses in areas with population density <44 persons/km2 were much more likely to require resource-intensive interactive geocoding than those in areas with >920 persons/km2 (odds ratio (OR) = 5.24; 95% confidence interval (CI) = 4.23, 6.49), as were addresses collected from participants during in-person interviews compared with mailed questionnaires (OR = 1.83; 95% CI = 1.59, 2.11). This study demonstrates that population density and address ascertainment method can influence automated geocoding results and that high success in address level geocoding is achievable for large-scale studies covering wide geographical areas.

The effect of administrative boundaries and geocoding error on cancer rates in California

Geocoding is often used to produce maps of disease rates from the diagnosis addresses of incident cases to assist with disease surveillance, prevention, and control. In this process, diagnosis addresses are converted into latitude/longitude pairs which are then aggregated to produce rates at varying geographic scales such as Census tracts, neighborhoods, cities, counties, and states. The specific techniques used within geocoding systems have an impact on where the output geocode is located and can therefore have an effect on the derivation of disease rates at different geographic aggregations. This paper investigates how county-level cancer rates are affected by the choice of interpolation method when case data are geocoded to the ZIP code level. Four commonly used areal unit interpolation techniques are applied and the output of each is used to compute crude county-level five-year incidence rates of all cancers in California. We found that the rates observed for 44 out of the 58 counties in California vary based on which interpolation method is used, with rates in some counties increasing by nearly 400% between interpolation methods.

A research agenda: does geocoding positional error matter in health GIS studies?

Until recently, little attention has been paid to geocoding positional accuracy and its impacts on accessibility measures; estimates of disease rates; findings of disease clustering; spatial prediction and modeling of health outcomes; and estimates of individual exposures based on geographic proximity to pollutant and pathogen sources. It is now clear that positional errors can result in flawed findings and poor public health decisions. Yet the current state-of-practice is to ignore geocoding positional uncertainty, primarily because of a lack of theory, methods and tools for quantifying, modeling, and adjusting for geocoding positional errors in health analysis. This paper proposes a research agenda to address this need. It summarizes the basics of the geocoding process, its assumptions, and empirical evidence describing the magnitude of geocoding positional error. An overview of the impacts of positional error in health analysis, including accessibility, disease clustering, exposure reconstruction, and spatial weights estimation is presented. The proposed research agenda addresses five key needs: (1) a lack of standardized, open-access geocoding resources for use in health research; (2) a lack of geocoding validation datasets that will allow the evaluation of alternative geocoding engines and procedures; (3) a lack of spatially explicit geocoding positional error models; (4) a lack of resources for assessing the sensitivity of spatial analysis results to geocoding positional error; (5) a lack of demonstration studies that illustrate the sensitivity of health policy decisions to geocoding positional error.

Investigating impacts of positional error on potential health care accessibility

Accessibility to health services at the local or community level is an effective approach to measuring health care delivery in various constituencies in Canada and the United States. GIS and spatial methods play an important role in measuring potential access to health services. The Three-Step Floating Catchment Area (3SFCA) method is a GIS based procedure developed to calculate potential (spatial) accessibility as a ratio of primary health care (PHC) providers to the surrounding population in urban settings. This method uses PHC provider locations in textual/address format supplied by local, regional, or national health authorities. An automated geocoding procedure is normally used to convert such addresses to a pair of geographic coordinates. The accuracy of geocoding depends on the type of reference data and the amount of value-added effort applied. This research investigates the success and accuracy of six geocoding methods as well as how geocoding error affects the 3SFCA method. ArcGIS software is used for geocoding and spatial accessibility estimation. Results will focus on two implications of geocoding: (1) the success and accuracy of different automated and value-added geocoding; and (2) the implications of these geocoding methods for GIS-based methods that generalise results based on location data.

Spatial error in geocoding physician location data from the AMA Physician Masterfile: implications for spatial accessibility analysis

The accuracy of geocoding hinges on the quality of address information that serves as input to the geocoding process; however errors associated with poor address quality are rarely studied. This paper examines spatial errors that arise due to incorrect address information with respect to physician location data in the United States. Studies of spatial accessibility to physicians in the U.S. typically rely on data from the American Medical Association's Physician Masterfile. These data are problematic because a substantial proportion of physicians only report a mailing address, which is often the physician's home (residential) location, rather than the address for the location where health care is provided. The incorrect geocoding of physicians' practice locations based on inappropriate address information results in a form of geocoding error that has not been widely analyzed. Using data for the Chicago metropolitan region, we analyze the extent and implications of geocoding error for measurement of spatial accessibility to primary care physicians. We geocode the locations of primary care physicians based on mailing addresses and office addresses. The spatial mismatch between the two is computed at the county, zip code and point location scales. Although mailing and office address locations are quite close for many physicians, they are far apart (>20 km) for a substantial minority. Kernel density estimation is used to characterize the spatial distribution of physicians based on office and mailing addresses and to identify areas of high spatial mismatch between the two. Errors are socially and geographically uneven, resulting in overestimation of physician supply in some high-income suburban communities, and underestimation in certain central city locations where health facilities are concentrated. The resulting errors affect local measures of spatial accessibility to primary care, biasing statistical analyses of the associations between spatial access to care and health outcomes.

Error propagation models to examine the effects of geocoding quality on spatial analysis of individual-level datasets

The quality of geocoding has received substantial attention in recent years. A synthesis of published studies shows that the positional errors of street geocoding are somewhat unique relative to those of other types of spatial data: (1) the magnitude of error varies strongly across urban-rural gradients; (2) the direction of error is not uniform, but strongly associated with the properties of local street segments; (3) the distribution of errors does not follow a normal distribution, but is highly skewed and characterized by a substantial number of very large error values; and (4) the magnitude of error is spatially autocorrelated and is related to properties of the reference data. This makes it difficult to employ analytic approaches or Monte Carlo simulations for error propagation modeling because these rely on generalized statistical characteristics. The current paper describes an alternative empirical approach to error propagation modeling for geocoded data and illustrates its implementation using three different case-studies of geocoded individual-level datasets. The first case-study consists of determining the land cover categories associated with geocoded addresses using a point-in-raster overlay. The second case-study consists of a local hotspot characterization using kernel density analysis of geocoded addresses. The third case-study consists of a spatial data aggregation using enumeration areas of varying spatial resolution. For each case-study a high quality reference scenario based on address points forms the basis for the analysis, which is then compared to the result of various street geocoding techniques. Results show that the unique nature of the positional error of street geocoding introduces substantial noise in the result of spatial analysis, including a substantial amount of bias for some analysis scenarios. This confirms findings from earlier studies, but expands these to a wider range of analytical techniques.

Geostatistical analysis of health data with different levels of spatial aggregation

This paper presents a geostatistical approach to combine two geographical sets of area-based data into the mapping of disease risk, with an application to the rate of prostate cancer late-stage diagnosis in North Florida. This methodology is used to combine individual-level data assigned to census tracts for confidentiality reasons with individual-level data that were allocated to ZIP codes because of incomplete geocoding. This form of binomial kriging, which accounts for the population size and shape of each geographical unit, can generate choropleth or isopleth risk maps that are all coherent through spatial aggregation. Incorporation of both types of areal data reduces the loss of information associated with incomplete geocoding, leading to maps of risk estimates that are globally less smooth and with smaller prediction error variance.

Development of spatial density maps based on geoprocessing web services: application to tuberculosis incidence in Barcelona, Spain

Background

Health professionals and authorities strive to cope with heterogeneous data, services, and statistical models to support decision making on public health. Sophisticated analysis and distributed processing capabilities over geocoded epidemiological data are seen as driving factors to speed up control and decision making in these health risk situations. In this context, recent Web technologies and standards-based web services deployed on geospatial information infrastructures have rapidly become an efficient way to access, share, process, and visualize geocoded health-related information.

Methods

Data used on this study is based on Tuberculosis (TB) cases registered in Barcelona city during 2009. Residential addresses are geocoded and loaded into a spatial database that acts as a backend database. The web-based application architecture and geoprocessing web services are designed according to the Representational State Transfer (REST) principles. These web processing services produce spatial density maps against the backend database.

Results

The results are focused on the use of the proposed web-based application to the analysis of TB cases in Barcelona. The application produces spatial density maps to ease the monitoring and decision making process by health professionals. We also include a discussion of how spatial density maps may be useful for health practitioners in such contexts.

Conclusions

In this paper, we developed web-based client application and a set of geoprocessing web services to support specific health-spatial requirements. Spatial density maps of TB incidence were generated to help health professionals in analysis and decision-making tasks. The combined use of geographic information tools, map viewers, and geoprocessing services leads to interesting possibilities in handling health data in a spatial manner. In particular, the use of spatial density maps has been effective to identify the most affected areas and its spatial impact. This study is an attempt to demonstrate how web processing services together with web-based mapping capabilities suit the needs of health practitioners in epidemiological analysis scenarios.

[Potential accessibility to mental health services in Montreal: a geographical information system approach]

BACKGROUND

The inequitable spatial distribution of health resources is a major public health problem because it exacerbates the spatial disparities of access and use of health services. However, efforts to quantify the problem and its consequences on public health have been hampered by a lack of adequate measures and methods. This study explores the spatial potential accessibility to mental health services in a heterogeneous urban environment and evaluates inequities in access to services in deprived areas.

METHODS

The study examines the spatial accessibility to mental health services in the Island of Montreal. All mental health services were geocoded from the six-digital postal code using the software ArcGIS 9.3. Accessibility was assessed through the two step floating catchment area method using the shortest route through a road network more often called reticular distance. This method takes into account the whole population, which is considered as the potential demand.

RESULTS

In general, accessibility to mental health services seems high in Montreal. It can be seen that at a distance of 1 km, nearly 90% of the territory is accessible. However, we also note that accessibility scores greatly diminish with distance. At 1 km, there are about 10.05 services for 10,000 persons and at 3 km, there is only 1.12 services for 10,000 persons. Over 50% of non-accessible areas are concentrated in the first quartile of deprivation and less than 10% are found in the fourth quartile, indicating good accessibility in severely deprived areas.

CONCLUSION

Accessibility to health services will always be the dominant issue debate in developing and undeveloped countries over the next decade. It is therefore urgent to develop technical and methodological tools to study and anticipate areas that may face services' shortage.

Optimizing the two-step floating catchment area method for measuring spatial accessibility to medical clinics in Montreal

BACKGROUND

Reducing spatial access disparities to healthcare services is a growing priority for healthcare planners especially among developed countries with aging populations. There is thus a pressing need to determine which populations do not enjoy access to healthcare, yet efforts to quantify such disparities in spatial accessibility have been hampered by a lack of satisfactory measurements and methods. This study compares an optimised and the conventional version of the two-step floating catchment area (2SFCA) method to assess spatial accessibility to medical clinics in Montreal.

METHODS

We first computed catchments around existing medical clinics of Montreal Island based on the shortest network distance. Population nested in dissemination areas were used to determine potential users of a given medical clinic. To optimize the method, medical clinics (supply) were weighted by the number of physicians working in each clinic, while the previous year's medical clinic users were computed by ten years age group was used as weighting coefficient for potential users of each medical clinic (demand).

RESULTS

The spatial accessibility score (SA) increased considerably with the optimisation method. Within a distance of 1 Km, for instance, the maximum clinic accessible for 1,000 persons is 2.4 when the conventional method is used, compared with 27.7 for the optimized method. The t-test indicates a significant difference between the conventional and the optimized 2SFCA methods. Also, results of the differences between the two methods reveal a clustering of residuals when distance increases. In other words, a low threshold would be associated with a lack of precision.

CONCLUSION

Results of this study suggest that a greater effort must be made ameliorate spatial accessibility to medical clinics in Montreal. To ensure that health resources are allocated in the interest of the population, health planners and the government should consider a strategy in the sitting of future clinics which would provide spatial access to the greatest number of people.

A geospatial analysis of the impact of sex offender residency restrictions in two New York counties

The efficacy of sex offender residence restriction laws in enhancing public safety is controversial and further complicated by evidence that adverse collateral effects may negate or even outweigh whatever benefits they achieve. Based on the theory of "distance decay" that postulates that offenders are more likely to recidivate closer to home, the statutes seek to distance offenders from potential child victims. However, to the extent that such statutes preclude residence in large portions of covered jurisdictions, it has been argued that they contribute to social instability, relegation of offenders to rural or undesirable locations, and even homelessness. A small number of studies have demonstrated the impact of restrictions on residential availability and compliance with the laws, but methodologic issues make it difficult to compare findings. This study uses parcel geocoding, a computerized mapping method, to examine the impact of the sex offender residency restrictions enacted in Erie and Schenectady Counties, NY. Identification and mapping of restricted locations revealed that in nonurban areas, available residential locations were drastically reduced by the restrictions (89.46% and 73.16% restricted in the two counties) and in urban areas almost completely eliminated (95.45% and 97.21%). Unexpectedly, however, when the registered sex offenders in each county were matched to their addresses in the state database, analysis revealed that residence restrictions had no demonstrable effect on where offenders live. More than 85% of offenders in each of the counties were found living in the urban centers, the vast majority of whom (91.89% and 100%) were matched to addresses in restricted locations. These findings may have important policy and procedural implications in the creation and enforcement of sex offender statutes, as well as in the evaluation of those presently in place.

Neighborhood-level built environment and social characteristics associated with serious childhood motor vehicle occupant injuries

The effect of residential neighborhood characteristics on a child's risk of serious motor vehicle traffic occupant injuries was evaluated in New York State, USA, for the years 1993-2003, with particular focus on the effect of neighborhood walkability. Risk increased significantly (p < 0.0001) with decreasing street connectivity and as more workers commuted more than 30 min using means other than public transportation, along with more single-parent households and less college attainment in the neighborhood, regardless of whether New York City was in the study. After adjusting for age, gender and socio-economic community factors, the apparent loss of walkability in a child's neighborhood increases their risk of serious injury as an occupant of a motor vehicle.

Geographical information system and environmental epidemiology: a cross-sectional spatial analysis of the effects of traffic-related air pollution on population respiratory health

BACKGROUND

Traffic-related air pollution is a potential risk factor for human respiratory health. A Geographical Information System (GIS) approach was used to examine whether distance from a main road (the Tosco-Romagnola road) affected respiratory health status.

METHODS

We used data collected during an epidemiological survey performed in the Pisa-Cascina area (central Italy) in the period 1991-93. A total of 2841 subjects participated in the survey and filled out a standardized questionnaire on health status, socio-demographic information, and personal habits. A variable proportion of subjects performed lung function and allergy tests. Highly exposed subjects were defined as those living within 100 m of the main road, moderately exposed as those living between 100 and 250 m from the road, and unexposed as those living between 250 and 800 m from the road. Statistical analyses were conducted to compare the risks for respiratory symptoms and diseases between exposed and unexposed. All analyses were stratified by gender.

RESULTS

The study comprised 2062 subjects: mean age was 45.9 years for men and 48.9 years for women. Compared to subjects living between 250 m and 800 m from the main road, subjects living within 100 m of the main road had increased adjusted risks for persistent wheeze (OR = 1.76, 95% CI = 1.08-2.87), COPD diagnosis (OR = 1.80, 95% CI = 1.03-3.08), and reduced FEV1/FVC ratio (OR = 2.07, 95% CI = 1.11-3.87) among males, and for dyspnea (OR = 1.61, 95% CI = 1.13-2.27), positivity to skin prick test (OR = 1.83, 95% CI = 1.11-3.00), asthma diagnosis (OR = 1.68, 95% CI = 0.97-2.88) and attacks of shortness of breath with wheeze (OR = 1.67, 95% CI = 0.98-2.84) among females.

CONCLUSION

This study points out the potential effects of traffic-related air pollution on respiratory health status, including lung function impairment. It also highlights the added value of GIS in environmental health research.

Accuracy of commercially available residential histories for epidemiologic studies

A key problem facing epidemiologists who wish to account for residential mobility in their analyses is the cost and difficulty of obtaining residential histories. Commercial residential history data of acceptable accuracy, cost, and coverage would be of great value. The present research evaluated the accuracy of residential histories from LexisNexis, Inc. The authors chose LexisNexis because the Michigan Cancer Registry has considered using their data, they have excellent procedures for privacy protection, and they make available residential histories at 25 cents per person. Only first and last name and address at last-known residence are required to access the residential history. The authors compared lifetime residential histories collected through the use of written surveys in a case-control study of bladder cancer in Michigan to the 3 residential addresses routinely available in the address history from LexisNexis. The LexisNexis address matches, as a whole, accounted for 71.5% of participants' lifetime addresses. These results provided a level of accuracy that indicates routine use of residential histories from commercial vendors is feasible. More detailed residential histories are available at a higher cost but were not analyzed in this study. Although higher accuracy is desirable, LexisNexis data are a vast improvement over the assumption of immobile individuals currently used in many spatial and spatiotemporal studies.

Geocoding rural addresses in a community contaminated by PFOA: a comparison of methods

BACKGROUND

Location is often an important component of exposure assessment, and positional errors in geocoding may result in exposure misclassification. In rural areas, successful geocoding to a street address is limited by rural route boxes. Communities have assigned physical street addresses to rural route boxes as part of E911 readdressing projects for improved emergency response. Our study compared automated and E911 methods for recovering and geocoding valid street addresses and assessed the impact of positional errors on exposure classification.

METHODS

The current study is a secondary analysis of existing data that included 135 addresses self-reported by participants of a rural community study who were exposed via public drinking water to perfluorooctanoate (PFOA) released from a DuPont facility in Parkersburg, West Virginia. We converted pre-E911 to post-E911 addresses using two methods: automated ZP4 address-correction software with the U.S. Postal Service LACS database and E911 data provided by Wood County, West Virginia. Addresses were geocoded using TeleAtlas, an online commercial service, and ArcView with StreetMap Premium North America NAVTEQ 2008 enhanced street dataset. We calculated positional errors using GPS measurements collected at each address and assessed exposure based on geocoded location in relation to public water pipes.

RESULTS

The county E911 data converted 89% of the eligible addresses compared to 35% by ZP4 LACS. ArcView/NAVTEQ geocoded more addresses (n = 130) and with smaller median distance between geocodes and GPS coordinates (39 meters) than TeleAtlas (n = 85, 188 meters). Without E911 address conversion, 25% of the geocodes would have been more than 1000 meters from the true location. Positional errors in TeleAtlas geocoding resulted in exposure misclassification of seven addresses whereas ArcView/NAVTEQ methods did not misclassify any addresses.

CONCLUSIONS

Although the study was limited by small numbers, our results suggest that the use of county E911 data in rural areas increases the rate of successful geocoding. Furthermore, positional accuracy of rural addresses in the study area appears to vary by geocoding method. In a large epidemiological study investigating the health effects of PFOA-contaminated public drinking water, this could potentially result in exposure misclassification if addresses are incorrectly geocoded to a street segment not serviced by public water.

Characterizing urban traffic exposures using transportation planning tools: an illustrated methodology for health researchers

Exposure to elevated levels of vehicular traffic has been associated with adverse cardiovascular and respiratory health effects in a range of populations, including children, the elderly, and individuals with pre-existing heart conditions, diabetes, obesity, and genetic susceptibilities. As these relationships become clearer, public health officials will need to have access to methods to identify areas of concern in terms of elevated traffic levels and susceptible populations. This paper briefly reviews current approaches for characterizing traffic exposure and then presents a detailed method that can be employed by public health officials and other researchers in performing screening assessments to define areas of potential concern within a particular locale and, with appropriate caveats, in epidemiologic studies examining traffic-related health impacts at the intra-urban scale. The method is based on two exposure parameters extensively used in numerous epidemiologic studies of traffic and health-proximity to high traffic roadways and overall traffic density. The method is demonstrated with publically available information on susceptible populations, traffic volumes, and Traffic Analysis Zones, a transportation planning tool long used by Metropolitan Planning Agencies and planners across the USA but presented here as a new application which can be used to spatially assess possible traffic-related impacts on susceptible populations. Recommendations are provided for the appropriate use of this methodology, along with its limitations.

The effects of local street network characteristics on the positional accuracy of automated geocoding for geographic health studies

Background

Automated geocoding of patient addresses for the purpose of conducting spatial epidemiologic studies results in positional errors. It is well documented that errors tend to be larger in rural areas than in cities, but possible effects of local characteristics of the street network, such as street intersection density and street length, on errors have not yet been documented. Our study quantifies effects of these local street network characteristics on the means and the entire probability distributions of positional errors, using regression methods and tolerance intervals/regions, for more than 6000 geocoded patient addresses from an Iowa county.

Results

Positional errors were determined for 6376 addresses in Carroll County, Iowa, as the vector difference between each 100%-matched automated geocode and its ground-truthed location. Mean positional error magnitude was inversely related to proximate street intersection density. This effect was statistically significant for both rural and municipal addresses, but more so for the former. Also, the effect of street segment length on geocoding accuracy was statistically significant for municipal, but not rural, addresses; for municipal addresses mean error magnitude increased with length.

Conclusion

Local street network characteristics may have statistically significant effects on geocoding accuracy in some places, but not others. Even in those locales where their effects are statistically significant, street network characteristics may explain a relatively small portion of the variability among geocoding errors. It appears that additional factors besides rurality and local street network characteristics affect accuracy in general.

Using imputation to provide location information for nongeocoded addresses

Background

The importance of geography as a source of variation in health research continues to receive sustained attention in the literature. The inclusion of geographic information in such research often begins by adding data to a map which is predicated by some knowledge of location. A precise level of spatial information is conventionally achieved through geocoding, the geographic information system (GIS) process of translating mailing address information to coordinates on a map. The geocoding process is not without its limitations, though, since there is always a percentage of addresses which cannot be converted successfully (nongeocodable). This raises concerns regarding bias since traditionally the practice has been to exclude nongeocoded data records from analysis.

Methodology/Principal Findings

In this manuscript we develop and evaluate a set of imputation strategies for dealing with missing spatial information from nongeocoded addresses. The strategies are developed assuming a known zip code with increasing use of collateral information, namely the spatial distribution of the population at risk. Strategies are evaluated using prostate cancer data obtained from the Maryland Cancer Registry. We consider total case enumerations at the Census county, tract, and block group level as the outcome of interest when applying and evaluating the methods. Multiple imputation is used to provide estimated total case counts based on complete data (geocodes plus imputed nongeocodes) with a measure of uncertainty. Results indicate that the imputation strategy based on using available population-based age, gender, and race information performed the best overall at the county, tract, and block group levels.

Conclusions/Significance

The procedure allows for the potentially biased and likely under reported outcome, case enumerations based on only the geocoded records, to be presented with a statistically adjusted count (imputed count) with a measure of uncertainty that are based on all the case data, the geocodes and imputed nongeocodes. Similar strategies can be applied in other analysis settings.

Spatial autocorrelation among automated geocoding errors and its effects on testing for disease clustering

Automated geocoding of patient addresses is an important data assimilation component of many spatial epidemiologic studies. Inevitably, the geocoding process results in positional errors. Positional errors incurred by automated geocoding tend to reduce the power of tests for disease clustering and otherwise affect spatial analytic methods. However, there are reasons to believe that the errors may often be positively spatially correlated and that this may mitigate their deleterious effects on spatial analyses. In this article, we demonstrate explicitly that the positional errors associated with automated geocoding of a data set of more than 6000 addresses in Carroll County, Iowa are spatially autocorrelated. Furthermore, through two simulation studies of disease processes, including one in which the disease process is overlain upon the Carroll County addresses, we show that spatial autocorrelation among geocoding errors maintains the power of two tests for disease clustering at a level higher than that which would occur if the errors were independent. Implications of these results for cluster detection, privacy protection, and measurement error modeling of geographic health data are discussed.

Enhancing spatial detection accuracy for syndromic surveillance with street level incidence data

Background

The Department of Defense Military Health System operates a syndromic surveillance system that monitors medical records at more than 450 non-combat Military Treatment Facilities (MTF) worldwide. The Electronic Surveillance System for Early Notification of Community-based Epidemics (ESSENCE) uses both temporal and spatial algorithms to detect disease outbreaks. This study focuses on spatial detection and attempts to improve the effectiveness of the ESSENCE implementation of the spatial scan statistic by increasing the spatial resolution of incidence data from zip codes to street address level.

Methods

Influenza-Like Illness (ILI) was used as a test syndrome to develop methods to improve the spatial accuracy of detected alerts. Simulated incident clusters of various sizes were superimposed on real ILI incidents from the 2008/2009 influenza season. Clusters were detected using the spatial scan statistic and their displacement from simulated loci was measured. Detected cluster size distributions were also evaluated for compliance with simulated cluster sizes.

Results

Relative to the ESSENCE zip code based method, clusters detected using street level incidents were displaced on average 65% less for 2 and 5 mile radius clusters and 31% less for 10 mile radius clusters. Detected cluster size distributions for the street address method were quasi normal and sizes tended to slightly exceed simulated radii. ESSENCE methods yielded fragmented distributions and had high rates of zero radius and oversized clusters.

Conclusions

Spatial detection accuracy improved notably with regard to both location and size when incidents were geocoded to street addresses rather than zip code centroids. Since street address geocoding success rates were only 73.5%, zip codes were still used for more than one quarter of ILI cases. Thus, further advances in spatial detection accuracy are dependant on systematic improvements in the collection of individual address information.

Evaluating geographic variation in type 1 and type 2 diabetes mellitus incidence in youth in four US regions

We evaluated geographic variation in type 1 and type 2 diabetes mellitus (T1DM, T2DM) in four regions of the United States. Data on 807 incident T1DM cases diabetes and 313 T2DM cases occurring in 2002-03 in South Carolina (SC) and Colorado (CO), 5 counties in Washington (WA), and an 8 county region around Cincinnati, Ohio (OH) among youth aged 10-19 years were obtained from the SEARCH for Diabetes in Youth Study. Geographic patterns were evaluated in a Bayesian framework. Incidence rates differed between the study regions, even within race/ethnic groups. Significant small-area variation within study region was observed for T1DM and T2DM. Evidence for joint spatial correlation between T1DM and T2DM was present at the county level for SC (r(SC)=0.31) and CO non-Hispanic Whites (r(CO)=0.40) and CO Hispanics (r(CO)=0.72). At the tract level, no evidence for meaningful joint spatial correlation was observed (r(SC)=-0.02; r(CO)=-0.02; r(OH)=0.03; and r(WA=)0.09). Our study provides evidence for the presence of both regional and small area, localized variation in type 1 and type 2 incidence among youth aged 10-19 years in the United States.

Assessment of community food resources: A Latino neighborhood study in upstate New York

This study aims to assess availability, affordability, and accessibility of food items in a low-income Latino neighborhood within a small city using an on-site food store survey. Store locations were identified by on-site GPS. Results showed the Latino neighborhood had limited availability and above average cost of high-fiber bread. Fresh vegetables were more expensive compared to the non-Latino neighborhood, and more stores in the Latino neighborhood participated in Supplemental Nutrition Assistance Food Program. The lack of supermarkets, fewer stores with disability access, and the lack of public transportation left Latino residents without a vehicle or with physical disabilities with few food shopping options.

Methods for retrospective geocoding in population studies: the Jackson Heart Study

The increasing use of geographic information systems (GIS) in epidemiological population studies requires careful attention to the methods employed in accomplishing geocoding and creating a GIS. Studies have provided limited details,hampering the ability to assess validity of spatial data. The purpose of this paper is to describe the multiphase geocoding methods used to retrospectively create a GIS in the Jackson Heart Study (JHS). We used baseline data from 5,302 participants enrolled in the JHS between 2000 and 2004 in a multiphase process to accomplish geocoding2 years after participant enrollment. After initial deletion of ungeocodable addresses(n=52), 96% were geocoded using ArcGIS. An interactive method using data abstraction from participant records, use of additional maps and street reference files,and verification of existence of address, yielded successful geocoding of all but 13 addresses. Overall, nearly 99% (n=5,237) of the JHS cohort was geocoded retrospectively using the multiple strategies for improving and locating geocodable addresses. Geocoding validation procedures revealed highly accurate and reliable geographic data. Using the methods and protocol developed provided a reliable spatial database that can be used for further investigation of spatial epidemiology. Baseline results were used to describe participants by select geographic indicators, including residence in urban or rural areas, as well as to validate the effectiveness of the study's sampling plan. Further, our results indicate that retrospectively developing a reliable GIS for a large, epidemiological study is feasible. This paper describes some of the challenges in retrospectively creating a GIS and provides practical tips that enhanced the success.

Birth and fetal death records and environmental exposures: promising data elements for environmental public health tracking of reproductive outcomes

OBJECTIVES

We inventoried and reviewed the birth and fetal death certificates of all 50 U.S. states to identify nonstandard data items that are environmentally relevant, inexpensive to collect, and might enhance environmental public health tracking.

METHODS

We obtained online or requested by mail or telephone the birth certificate and fetal death record forms or formats from each state. Every state data element was compared to the 2003 standards promulgated by the National Center for Health Statistics to identify any items that are not included on the standard. We then evaluated these items for their utility in environmentally related analyses.

RESULTS

We found three data fields of potential interest. First, although every state included residence of mother at time of delivery on the birth certificate, only four states collected information on how long the mother had lived there. This item may be useful in that it could be used to assess and reduce misclassification of environmental exposures among women during pregnancy. Second, we found that father's address was listed on the birth certificates of eight states. This data field may be useful for defining paternal environmental exposures, especially in cases where the parents do not live together. Third, parental occupation was listed on the birth certificates of 15 states and may be useful for defining parental workplace exposures. Our findings were similar for fetal death records.

CONCLUSION

If these data elements are accurate and well-reported, their addition to birth, fetal death, and other health records may aid in environmental public health tracking.

Local indicators of geocoding accuracy (LIGA): theory and application

BACKGROUND

Although sources of positional error in geographic locations (e.g. geocoding error) used for describing and modeling spatial patterns are widely acknowledged, research on how such error impacts the statistical results has been limited. In this paper we explore techniques for quantifying the perturbability of spatial weights to different specifications of positional error.

RESULTS

We find that a family of curves describes the relationship between perturbability and positional error, and use these curves to evaluate sensitivity of alternative spatial weight specifications to positional error both globally (when all locations are considered simultaneously) and locally (to identify those locations that would benefit most from increased geocoding accuracy). We evaluate the approach in simulation studies, and demonstrate it using a case-control study of bladder cancer in south-eastern Michigan.

CONCLUSION

Three results are significant. First, the shape of the probability distributions of positional error (e.g. circular, elliptical, cross) has little impact on the perturbability of spatial weights, which instead depends on the mean positional error. Second, our methodology allows researchers to evaluate the sensitivity of spatial statistics to positional accuracy for specific geographies. This has substantial practical implications since it makes possible routine sensitivity analysis of spatial statistics to positional error arising in geocoded street addresses, global positioning systems, LIDAR and other geographic data. Third, those locations with high perturbability (most sensitive to positional error) and high leverage (that contribute the most to the spatial weight being considered) will benefit the most from increased positional accuracy. These are rapidly identified using a new visualization tool we call the LIGA scatterplot.Herein lies a paradox for spatial analysis: For a given level of positional error increasing sample density to more accurately follow the underlying population distribution increases perturbability and introduces error into the spatial weights matrix. In some studies positional error may not impact the statistical results, and in others it might invalidate the results. We therefore must understand the relationships between positional accuracy and the perturbability of the spatial weights in order to have confidence in a study's results.

Evaluating geographic imputation approaches for zip code level data: an application to a study of pediatric diabetes

BACKGROUND

There is increasing interest in the study of place effects on health, facilitated in part by geographic information systems. Incomplete or missing address information reduces geocoding success. Several geographic imputation methods have been suggested to overcome this limitation. Accuracy evaluation of these methods can be focused at the level of individuals and at higher group-levels (e.g., spatial distribution).

METHODS

We evaluated the accuracy of eight geo-imputation methods for address allocation from ZIP codes to census tracts at the individual and group level. The spatial apportioning approaches underlying the imputation methods included four fixed (deterministic) and four random (stochastic) allocation methods using land area, total population, population under age 20, and race/ethnicity as weighting factors. Data included more than 2,000 geocoded cases of diabetes mellitus among youth aged 0-19 in four U.S. regions. The imputed distribution of cases across tracts was compared to the true distribution using a chi-squared statistic.

RESULTS

At the individual level, population-weighted (total or under age 20) fixed allocation showed the greatest level of accuracy, with correct census tract assignments averaging 30.01% across all regions, followed by the race/ethnicity-weighted random method (23.83%). The true distribution of cases across census tracts was that 58.2% of tracts exhibited no cases, 26.2% had one case, 9.5% had two cases, and less than 3% had three or more. This distribution was best captured by random allocation methods, with no significant differences (p-value > 0.90). However, significant differences in distributions based on fixed allocation methods were found (p-value < 0.0003).

CONCLUSION

Fixed imputation methods seemed to yield greatest accuracy at the individual level, suggesting use for studies on area-level environmental exposures. Fixed methods result in artificial clusters in single census tracts. For studies focusing on spatial distribution of disease, random methods seemed superior, as they most closely replicated the true spatial distribution. When selecting an imputation approach, researchers should consider carefully the study aims.

Neighbourhood deprivation and access to alcohol outlets: a national study

People living in poor areas suffer higher mortality than those in wealthy areas. Environmental factors partly explain this association, including exposure to pollutants and accessibility of healthcare. We sought to determine whether proximity to alcohol outlets varied by area deprivation in New Zealand. Roadway travel distance from each census unit to the nearest alcohol outlet was summarised according to socioeconomic deprivation for each area. Analyses were conducted by license type (pubs/bars, clubs, restaurants, off-licenses) and community urban-rural status. Strong associations were found between proximity to the nearest alcohol outlet and deprivation, there being greater access to outlets in more-deprived urban areas.

Development of an interactive environmental public health tracking system for data analysis, visualization, and reporting

Healthcare providers and governmental agencies routinely collect and report data on health outcomes. In addition, governmental agencies and industry collect and report information on environmental hazards and exposures that may impact health. Use of these data for environmental public health tracking has been a challenge because these data are managed by different data stewards, may contain confidential information that must be protected, and have not been collected in a manner to facilitate linkages. Available tools for analysis, visualization, and reporting of these data are either difficult to use or not available through a common user interface. The New York State Department of Health has developed a user-friendly interactive system to access and link these data while protecting confidential information. The Environmental Public health tracking system provides tools for describing the geographic patterns, trends, and statistical associations between health, environmental exposure, and environmental hazard data. These tools provide descriptive statistics and automated techniques that smooth the data in order to protect patient confidentiality and reduce random fluctuations in rates due to small numbers. This article describes the user interface, data linkages, and analytic, visualization, and reporting tools.

Neighborhood environments: disparities in access to healthy foods in the U.S

BACKGROUND

Poor dietary patterns and obesity, established risk factors for chronic disease, have been linked to neighborhood deprivation, neighborhood minority composition, and low area population density. Neighborhood differences in access to food may have an important influence on these relationships and health disparities in the U.S. This article reviews research relating to the presence, nature, and implications of neighborhood differences in access to food.

METHODS

A snowball strategy was used to identify relevant research studies (n=54) completed in the U.S. and published between 1985 and April 2008.

RESULTS

Research suggests that neighborhood residents who have better access to supermarkets and limited access to convenience stores tend to have healthier diets and lower levels of obesity. Results from studies examining the accessibility of restaurants are less consistent, but there is some evidence to suggest that residents with limited access to fast-food restaurants have healthier diets and lower levels of obesity. National and local studies across the U.S. suggest that residents of low-income, minority, and rural neighborhoods are most often affected by poor access to supermarkets and healthful food. In contrast, the availability of fast-food restaurants and energy-dense foods has been found to be greater in lower-income and minority neighborhoods.

CONCLUSIONS

Neighborhood disparities in access to food are of great concern because of their potential to influence dietary intake and obesity. Additional research is needed to address various limitations of current studies, identify effective policy actions, and evaluate intervention strategies designed to promote more equitable access to healthy foods.

An effective and efficient approach for manually improving geocoded data

Background

The process of geocoding produces output coordinates of varying degrees of quality. Previous studies have revealed that simply excluding records with low-quality geocodes from analysis can introduce significant bias, but depending on the number and severity of the inaccuracies, their inclusion may also lead to bias. Little quantitative research has been presented on the cost and/or effectiveness of correcting geocodes through manual interactive processes, so the most cost effective methods for improving geocoded data are unclear. The present work investigates the time and effort required to correct geocodes contained in five health-related datasets that represent examples of data commonly used in Health GIS.

Results

Geocode correction was attempted on five health-related datasets containing a total of 22,317 records. The complete processing of these data took 11.4 weeks (427 hours), averaging 69 seconds of processing time per record. Overall, the geocodes associated with 12,280 (55%) of records were successfully improved, taking 95 seconds of processing time per corrected record on average across all five datasets. Geocode correction improved the overall match rate (the number of successful matches out of the total attempted) from 79.3 to 95%. The spatial shift between the location of original successfully matched geocodes and their corrected improved counterparts averaged 9.9 km per corrected record. After geocode correction the number of city and USPS ZIP code accuracy geocodes were reduced from 10,959 and 1,031 to 6,284 and 200, respectively, while the number of building centroid accuracy geocodes increased from 0 to 2,261.

Conclusion

The results indicate that manual geocode correction using a web-based interactive approach is a feasible and cost effective method for improving the quality of geocoded data. The level of effort required varies depending on the type of data geocoded. These results can be used to choose between data improvement options (e.g., manual intervention, pseudocoding/geo-imputation, field GPS readings).

Children at risk: measuring racial/ethnic disparities in potential exposure to air pollution at school and home

STUDY OBJECTIVE

This paper addresses the environmental justice implications of children's health by exploring racial/ethnic disparities in potential exposure to air pollution, based on both school and home locations of children and three different types of pollution sources, in Orange County, Florida, USA.

METHODS

Using geocoded school and residence locations of 151 709 children enrolled in the public school system, distribution functions of proximity to the nearest source are generated for each type of air pollution source in order to compare the exposure potential of white, Hispanic, and black children. Discrete buffer distances are utilised to provide quantitative comparisons for statistical testing.

MAIN RESULTS

At any given distance from each type of pollution source, the cumulative proportion of Hispanic or black children significantly exceeds the corresponding proportion of white children, for both school and home locations. Regardless of race, however, a larger proportion of children are potentially exposed to air pollution at home than at school.

CONCLUSIONS

This study addresses the growing need to consider both daytime and nighttime activity patterns in the assessment of children's exposure to environmental hazards and related health risks. The results indicate a consistent pattern of racial inequity in the spatial distribution of all types of air pollution sources examined, with black children facing the highest relative levels of potential exposure at both school and home locations.

Factors associated with endemic raccoon (Procyon lotor) rabies in terrestrial mammals in New York State, USA

This study evaluated characteristics associated with raccoon (Procyon lotor) rabies in New York State (NYS), USA, where this disease has been endemic for the last 15 years. The study included 4448 cases of raccoon rabies in terrestrial mammals reported across 1639 census tracts of NYS during 1997-2003. A Poisson-regression model with census tract-year as the unit of analysis revealed a higher number of raccoon-variant rabies cases per square kilometer in census tracts with each percent increase in the proportion of low-intensity residential areas (those with a lower concentration of housing units) (RR=7.68) and a lack of rivers/lakes (RR=1.20) and major roads (RR=1.10), while the number of cases decreased with each 1-m increase in land elevation (RR=0.998), and each percent increase in the proportion of wetlands (RR=0.01). The model was adjusted for county, ecoregion, and latitude to help control for unknown spatially dependent covariates. The model may be used in prioritizing areas for rabies control based on differential risk, including use of costly intervention methods such as oral rabies vaccine.

Geocoding accuracy and the recovery of relationships between environmental exposures and health

Background

This research develops methods for determining the effect of geocoding quality on relationships between environmental exposures and health. The likelihood of detecting an existing relationship – statistical power – between measures of environmental exposures and health depends not only on the strength of the relationship but also on the level of positional accuracy and completeness of the geocodes from which the measures of environmental exposure are made. This paper summarizes the results of simulation studies conducted to examine the impact of inaccuracies of geocoded addresses generated by three types of geocoding processes: a) addresses located on orthophoto maps, b) addresses matched to TIGER files (U.S Census or their derivative street files); and, c) addresses from E-911 geocodes (developed by local authorities for emergency dispatch purposes).

Results

The simulated odds of disease using exposures modelled from the highest quality geocodes could be sufficiently recovered using other, more commonly used, geocoding processes such as TIGER and E-911; however, the strength of the odds relationship between disease exposures modelled at geocodes generally declined with decreasing geocoding accuracy.

Conclusion

Although these specific results cannot be generalized to new situations, the methods used to determine the sensitivity of results can be used in new situations. Estimated measures of positional accuracy must be used in the interpretation of results of analyses that investigate relationships between health outcomes and exposures measured at residential locations. Analyses similar to those employed in this paper can be used to validate interpretation of results from empirical analyses that use geocoded locations with estimated measures of positional accuracy.

Using built environment characteristics to predict walking for exercise

Background

Environments conducive to walking may help people avoid sedentary lifestyles and associated diseases. Recent studies developed walkability models combining several built environment characteristics to optimally predict walking. Developing and testing such models with the same data could lead to overestimating one's ability to predict walking in an independent sample of the population. More accurate estimates of model fit can be obtained by splitting a single study population into training and validation sets (holdout approach) or through developing and evaluating models in different populations. We used these two approaches to test whether built environment characteristics near the home predict walking for exercise. Study participants lived in western Washington State and were adult members of a health maintenance organization. The physical activity data used in this study were collected by telephone interview and were selected for their relevance to cardiovascular disease. In order to limit confounding by prior health conditions, the sample was restricted to participants in good self-reported health and without a documented history of cardiovascular disease.

Results

For 1,608 participants meeting the inclusion criteria, the mean age was 64 years, 90 percent were white, 37 percent had a college degree, and 62 percent of participants reported that they walked for exercise. Single built environment characteristics, such as residential density or connectivity, did not significantly predict walking for exercise. Regression models using multiple built environment characteristics to predict walking were not successful at predicting walking for exercise in an independent population sample. In the validation set, none of the logistic models had a C-statistic confidence interval excluding the null value of 0.5, and none of the linear models explained more than one percent of the variance in time spent walking for exercise. We did not detect significant differences in walking for exercise among census areas or postal codes, which were used as proxies for neighborhoods.

Conclusion

None of the built environment characteristics significantly predicted walking for exercise, nor did combinations of these characteristics predict walking for exercise when tested using a holdout approach. These results reflect a lack of neighborhood-level variation in walking for exercise for the population studied.

Validation of a GIS facilities database: quantification and implications of error

PURPOSE

To validate a commercial database of community-level physical activity facilities that can be used in future research examining associations between access to physical activity facilities and individual-level physical activity and obesity.

METHODS

Physical activity facility characteristics and locations obtained from a commercial database were compared to a field census conducted in 80 census block groups within two U.S. communities. Agreement statistics, agreement of administratively defined neighborhoods, and distance between locations were used to quantify count, attribute, and positional error.

RESULTS

There was moderate agreement (concordance: nonurban: 0.39; urban: 0.46) of presence of any physical activity facility and poor to moderate agreement (kappa range: 0.14 to 0.76) of physical activity facility type. The mean Euclidean distance between commercial database versus field census locations was 757 and 35 m in the nonurban and urban communities, respectively. However, 94% and 100% of nonurban and urban physical activity facilities, respectively, fell into the same 5-digit ZIP code, dropping to 92% and 98% in the same block group and 71% along the same street.

CONCLUSIONS

Our findings suggest that the commercial database of physical activity facilities may contain appreciable error, but patterns of error suggest that built environment-health associations are likely biased downward.

Estimating the accuracy of geographical imputation

BACKGROUND

To reduce the number of non-geocoded cases researchers and organizations sometimes include cases geocoded to postal code centroids along with cases geocoded with the greater precision of a full street address. Some analysts then use the postal code to assign information to the cases from finer-level geographies such as a census tract. Assignment is commonly completed using either a postal centroid or by a geographical imputation method which assigns a location by using both the demographic characteristics of the case and the population characteristics of the postal delivery area. To date no systematic evaluation of geographical imputation methods ("geo-imputation") has been completed. The objective of this study was to determine the accuracy of census tract assignment using geo-imputation.

METHODS

Using a large dataset of breast, prostate and colorectal cancer cases reported to the New Jersey Cancer Registry, we determined how often cases were assigned to the correct census tract using alternate strategies of demographic based geo-imputation, and using assignments obtained from postal code centroids. Assignment accuracy was measured by comparing the tract assigned with the tract originally identified from the full street address.

RESULTS

Assigning cases to census tracts using the race/ethnicity population distribution within a postal code resulted in more correctly assigned cases than when using postal code centroids. The addition of age characteristics increased the match rates even further. Match rates were highly dependent on both the geographic distribution of race/ethnicity groups and population density.

CONCLUSION

Geo-imputation appears to offer some advantages and no serious drawbacks as compared with the alternative of assigning cases to census tracts based on postal code centroids. For a specific analysis, researchers will still need to consider the potential impact of geocoding quality on their results and evaluate the possibility that it might introduce geographical bias.