On the wrong side of the tracts? Evaluating the accuracy of geocoding in public health research

Enabling Hotspot Detection and Public Health Response to the COVID-19 Pandemic

Introduction

Public-facing maps of COVID-19 cases, hospital admissions, and deaths are commonly displayed at the state, county, and zip code levels, and low case counts are suppressed to protect confidentiality. Public health authorities are tasked with case identification, contact tracing, and canvasing for educational purposes during a pandemic. Given limited resources, authorities would benefit from the ability to tailor their efforts to a particular neighborhood or congregate living facility.

Methods

We describe the methods of building a real-time visualization of patients with COVID-19–positive tests, which facilitates timely public health response to the pandemic. We developed an interactive street-level visualization that shows new cases developing over time and resolving after 14 days of infection. Our source data included patient demographics (ie, age, race and ethnicity, and sex), street address of residence, respiratory test results, and date of test.

Results

We used colored dots to represent infections. The resulting animation shows where new cases developed in the region and how patterns changed over the course of the pandemic. Users can enlarge specific areas of the map and see street-level detail on residential location of each case and can select from demographic overlays and contour mapping options to see high-level patterns and associations with demographics and chronic disease prevalence as they emerge.

Conclusions

Before the development of this tool, local public health departments in our region did not have a means to map cases of disease to the street level and gain real-time insights into the underlying population where hotspots had developed. For privacy reasons, this tool is password-protected and not available to the public. We expect this tool to prove useful to public health departments as they navigate not only COVID-19 pandemic outcomes but also other public health threats, including chronic diseases and communicable disease outbreaks.

High-Risk Neighborhoods and Neurodevelopmental Outcomes in Infants Born Preterm

OBJECTIVE

To study the association between neighborhood risk and moderate to severe neurodevelopmental impairment (NDI) at 22-26 months corrected age in children born at <34 weeks of gestation. We hypothesized that infants born preterm living in high-risk neighborhoods would have a greater risk of NDI and cognitive, motor, and language delays.

STUDY DESIGN

We studied a retrospective cohort of 1291 infants born preterm between 2005 and 2016, excluding infants with congenital anomalies. NDI was defined as any one of the following: a Bayley Scales of Infant and Toddler Development-III Cognitive or Motor composite score <85, bilateral blindness, bilateral hearing impairment, or moderate-severe cerebral palsy. Maternal addresses were geocoded to identify census block groups and create high-risk versus low-risk neighborhood groups. Bivariate and regression analyses were run to assess the impact of neighborhood risk on outcomes.

RESULTS

Infants from high-risk (n = 538; 42%) and low-risk (n = 753; 58%) neighborhoods were compared. In bivariate analyses, the risk of NDI and cognitive, motor, and language delays was greater in high-risk neighborhoods. In adjusted regression models, the risks of NDI (OR, 1.43; 95% CI, 1.04-1.98), cognitive delay (OR, 1.62; 95% CI, 1.15-2.28), and language delay (OR, 1.58; 95% CI, 1.15-2.16) were greater in high-risk neighborhoods. Breast milk at discharge was more common in low-risk neighborhoods and was protective of NDI in regression analysis.

CONCLUSIONS

High neighborhood risk provides an independent contribution to preterm adverse NDI, cognitive, and language outcomes. In addition, breast milk at discharge was protective. Knowledge of neighborhood risk may inform the targeted implementation of programs for socially disadvantaged infants.

Geocoding precision of birth records from 2008 to 2017 in Kentucky, USA

Maternal address information captured on birth records is increasingly used to estimate residential environmental exposures during pregnancy. However, there has been limited assessment of the geocoding precision of birth records, particularly since the adoption of the 2003 standard birth certificate in 2015. To address this gap, this study evaluated the geocoding precision of live and stillbirth records of Kentucky residents over ten years, from 2008 through 2017. This study summarized the demographic characteristics of imprecisely geocoded records and, using a bivariate logistic regression, identified covariates associated with poor geocoding precision among three population density designations-metro, non-metro, and rural. We found that in metro areas, after adjusting for area deprivation, education, and the race, age and education of both parents, records for Black mothers had 48% lower odds of imprecise geocoding (aOR=0.52, 95% CI: 0.48, 0.56), while Black women in rural areas had 96% higher odds of imprecise geocoding (aOr=1.96, 95% CI: 1.68, 2.28). This study also found that over the study period, rural and non-metro areas began with a high proportion of imprecisely geocoded records (38% in rural areas, 19% in non-metro), but both experienced an 8% decline in imprecisely geocoded records over the study period (aOr=0.92, 95% CI: 0.92, 0.94). This study shows that, while geocoding precision has improved in Kentucky, further work is needed to improve geocoding in rural areas and address racial and ethnic disparities.

Quality of automatic geocoding tools: a study using addresses from hospital record files in Temuco, Chile

Abstract: Automatic geocoding methods have become popular in recent years, facilitating the study of the association between health outcomes and the place of living. However, rather few studies have evaluated geocoding quality, with most of them being performed in the US and Europe. This article aims to compare the quality of three automatic online geocoding tools against a reference method. A subsample of 300 handwritten addresses from hospital records was geocoded using Bing, Google Earth, and Google Maps. Match rates were higher (> 80%) for Google Maps and Google Earth compared with Bing. However, the accuracy of the addresses was better for Bing with a larger proportion (> 70%) of addresses with positional errors below 20m. Generally, performance did not vary for each method for different socioeconomic status. Overall, the methods showed an acceptable, but heterogeneous performance, which may be a warning against the use of automatic methods without assessing quality in other municipalities, particularly in Chile and Latin America.

Spatial Heterogeneity in Positional Errors: A Comparison of Two Residential Geocoding Efforts in the Agricultural Health Study

Geocoding is a powerful tool for environmental exposure assessments that rely on spatial databases. Geocoding processes, locators, and reference datasets have improved over time; however, improvements have not been well-characterized. Enrollment addresses for the Agricultural Health Study, a cohort of pesticide applicators and their spouses in Iowa (IA) and North Carolina (NC), were geocoded in 2012–2016 and then again in 2019. We calculated distances between geocodes in the two periods. For a subset, we computed positional errors using “gold standard” rooftop coordinates (IA; N = 3566) or Global Positioning Systems (GPS) (IA and NC; N = 1258) and compared errors between periods. We used linear regression to model the change in positional error between time periods (improvement) by rural status and population density, and we used spatial relative risk functions to identify areas with significant improvement. Median improvement between time periods in IA was 41 m (interquartile range, IQR: −2 to 168) and 9 m (IQR: −80 to 133) based on rooftop coordinates and GPS, respectively. Median improvement in NC was 42 m (IQR: −1 to 109 m) based on GPS. Positional error was greater in rural and low-density areas compared to in towns and more densely populated areas. Areas of significant improvement in accuracy were identified and mapped across both states. Our findings underscore the importance of evaluating determinants and spatial distributions of errors in geocodes used in environmental epidemiology studies.

Healthy Food Density is Not Associated With Diet Quality Among Pregnant Women With Overweight/Obesity in South Carolina

OBJECTIVE

Examine the association and moderating effect of residential location (urban/rural) on the relationship between neighborhood healthy food density and diet quality.

DESIGN

Cross-sectional analysis of baseline data from the Health in Pregnancy and Postpartum study, a randomized trial designed to prevent excessive gestational weight gain.

PARTICIPANTS

Pregnant women in South Carolina with prepregnancy overweight/obesity (n = 228).

MAIN OUTCOME MEASURES

Healthy Eating Index-2015 (HEI) was used to measure diet quality from 2 24-hour dietary recalls. The HEI total scores and 11 binary HEI components (those that met the standard for maximum component score vs those that did not) were calculated as dependent variables.

ANALYSIS

Multiple linear and logistic regression models were used to examine the association between healthy food density and HEI total scores and meeting the standards for maximum component scores. Healthy food density × residential location tested for moderation. P < 0.05 indicated significance.

RESULTS

Participants' diet quality was suboptimal (mean, 52.0; SD, 11.7; range, 27-85). Healthy food density was not significantly related to HEI total scores or components, and residential location was not a moderator.

CONCLUSIONS AND IMPLICATIONS

Diet quality was suboptimal, and there was no relationship between healthy food density and diet quality among Health in Pregnancy and Postpartum study participants. These data support examining behavioral factors that could influence diet quality.

Starter Kit for Geotagging and Geovisualization in Health Care: Resource Paper

BACKGROUND

Geotagging is the process of attaching geospatial tags to various media data types. In health care, the goal of geotagging is to gain a better understanding of health-related questions applied to populations. Although there has been a prevalence of geographic information in public health, in order to effectively use and expand geotagging across health care there is a requirement to understand other factors such as the disposition, standardization, data sources, technologies, and limitations.

OBJECTIVE

The objective of this document is to serve as a resource for new researchers in the field. This report aims to be comprehensive but easy for beginners to understand and adopt in practice. The optimal geocodes, their sources, and a rationale for use are suggested. Geotagging's issues and limitations are also discussed.

METHODS

A comprehensive review of technical instructions and articles was conducted to evaluate guidelines for geotagging, and online resources were curated to support the implementation of geotagging practices. Summary tables were developed to describe the available geotagging resources (free and for fee) that can be leveraged by researchers and quality improvement personnel to effectively perform geospatial analyses primarily targeting US health care.

RESULTS

This paper demonstrated steps to develop an initial geotagging and geovisualization project with clear structure and instructions. The geotagging resources were summarized. These resources are essential for geotagging health care projects. The discussion section provides better understanding of geotagging's limitations and suggests suitable way to approach it.

CONCLUSIONS

We explain how geotagging can be leveraged in health care and offer the necessary initial resources to obtain geocodes, adjustment data, and health-related measures. The resources outlined in this paper can support an individual and/or organization in initiating a geotagging health care project.

Racial Segregation Across U.S. Nursing Homes: A Systematic Review of Measurement and Outcomes

BACKGROUND AND OBJECTIVES

Nursing homes remain subjected to institutional racial segregation in the United States. However, a standardized approach to measure segregation in nursing homes does not appear to be established. A systematic review was conducted to identify all formal measurement approaches to evaluate racial segregation among nursing home facilities, and to then identify the association between segregation and quality of care in this context.

RESEARCH DESIGN AND METHODS

PubMed, Scopus, and Web of Science databases were searched (January 2018) for publications relating to nursing home segregation. Following the PRISMA guidelines, studies were included that formally measured racial segregation of nursing homes residents across facilities with regional-level data.

RESULTS

Eight studies met the inclusion criteria. Formal segregation measures included the Dissimilarity Index, Disparities Quality Index, Modified Thiel's Entropy Index, Gini coefficient, and adapted models. The most common data sources were the Minimum Data Set (MDS; resident-level), the Certification and Survey Provider Enhanced Reporting data (CASPER; facility-level), and the Area Resource File/ U.S. Census Data (regional-level). Most studies showed evidence of racial segregation among U.S. nursing home facilities and documented a negative impact of segregation on racial minorities and facility-level quality outcomes.

DISCUSSION AND IMPLICATIONS

The measurement of racial segregation among nursing homes is heterogeneous. While there are limitations to each methodology, this review can be used as a reference when trying to determine the best approach to measure racial segregation in future studies. Moreover, racial segregation among nursing homes remains a problem and should be further evaluated.

Lessons Learned From the Environmental Public Health Tracking Sub-County Data Pilot Project

OBJECTIVE

Small area data are key to better understanding the complex relationships between environmental health, health outcomes, and risk factors at a local level. In 2014, the Centers for Disease Control and Prevention's National Environmental Public Health Tracking Program (Tracking Program) conducted the Sub-County Data Pilot Project with grantees to consider integration of sub-county data into the National Environmental Public Health Tracking Network (Tracking Network).

DESIGN

The Tracking Program and grantees developed sub-county-level data for several data sets during this pilot project, working to standardize processes for submitting data and creating required geographies. Grantees documented challenges they encountered during the pilot project and documented decisions.

RESULTS

This article covers the challenges revealed during the project. It includes insights into geocoding, aggregation, population estimates, and data stability and provides recommendations for moving forward.

CONCLUSION

National standards for generating, analyzing, and sharing sub-county data should be established to build a system of sub-county data that allow for comparison of outcomes, geographies, and time. Increasing the availability and accessibility of small area data will not only enhance the Tracking Network's capabilities but also contribute to an improved understanding of environmental health and informed decision making at a local level.

GIScience and cancer: State of the art and trends for cancer surveillance and epidemiology

Maps are well recognized as an effective means of presenting and communicating health data, such as cancer incidence and mortality rates. These data can be linked to geographic features like counties or census tracts and their associated attributes for mapping and analysis. Such visualization and analysis provide insights regarding the geographic distribution of cancer and can be important for advancing effective cancer prevention and control programs. Applying a spatial approach allows users to identify location-based patterns and trends related to risk factors, health outcomes, and population health. Geographic information science (GIScience) is the discipline that applies Geographic Information Systems (GIS) and other spatial concepts and methods in research. This review explores the current state and evolution of GIScience in cancer research by addressing fundamental topics and issues regarding spatial data and analysis that need to be considered. GIScience, along with its health-specific application in the spatial epidemiology of cancer, incorporates multiple geographic perspectives pertaining to the individual, the health care infrastructure, and the environment. Challenges addressing these perspectives and the synergies among them can be explored through GIScience methods and associated technologies as integral parts of epidemiologic research, analysis efforts, and solutions. The authors suggest GIScience is a powerful tool for cancer research, bringing additional context to cancer data analysis and potentially informing decision-making and policy, ultimately aimed at reducing the burden of cancer.

Incorporating a location-based socioeconomic index into a de-identified i2b2 clinical data warehouse

OBJECTIVE

Clinical research data warehouses are largely populated from information extracted from electronic health records (EHRs). While these data provide information about a patient's medications, laboratory results, diagnoses, and history, her social, economic, and environmental determinants of health are also major contributing factors in readmission, morbidity, and mortality and are often absent or unstructured in the EHR. Details about a patient's socioeconomic status may be found in the U.S. census. To facilitate researching the impacts of socioeconomic status on health outcomes, clinical and socioeconomic data must be linked in a repository in a fashion that supports seamless interrogation of these diverse data elements. This study demonstrates a method for linking clinical and location-based data and querying these data in a de-identified data warehouse using Informatics for Integrating Biology and the Bedside.

MATERIALS AND METHODS

Patient data were extracted from the EHR at Nebraska Medicine. Socioeconomic variables originated from the 2011-2015 five-year block group estimates from the American Community Survey. Data querying was performed using Informatics for Integrating Biology and the Bedside. All location-based data were truncated to prevent identification of a location with a population <20 000 individuals.

RESULTS

We successfully linked location-based and clinical data in a de-identified data warehouse and demonstrated its utility with a sample use case.

DISCUSSION

With location-based data available for querying, research investigating the impact of socioeconomic context on health outcomes is possible. Efforts to improve geocoding can readily be incorporated into this model.

CONCLUSION

This study demonstrates a means for incorporating and querying census data in a de-identified clinical data warehouse.

A Community Needs Index for Adolescent Pregnancy Prevention Program Planning: Application of Spatial Generalized Linear Mixed Models

Objective The objective is to estimate community needs with respect to risky adolescent sexual behavior in a way that is risk-adjusted for multiple community factors. Methods Generalized linear mixed modeling was applied for estimating teen pregnancy and sexually transmitted disease (STD) incidence by postal ZIP code in New York State, in a way that adjusts for other community covariables and residual spatial autocorrelation. A community needs index was then obtained by summing the risk-adjusted estimates of pregnancy and STD cases. Results Poisson regression with a spatial random effect was chosen among competing modeling approaches. Both the risk-adjusted caseloads and rates were computed for ZIP codes, which allowed risk-based prioritization to help guide funding decisions for a comprehensive adolescent pregnancy prevention program. Conclusions This approach provides quantitative evidence of community needs with respect to risky adolescent sexual behavior, while adjusting for other community-level variables and stabilizing estimates in areas with small populations. Therefore, it was well accepted by the affected groups and proved valuable for program planning. This methodology may also prove valuable for follow up program evaluation. Current research is directed towards further improving the statistical modeling approach and applying to different health and behavioral outcomes, along with different predictor variables.

Individual, peer, and family factor modification of neighborhood-level effects on adolescent alcohol, cigarette, e-cigarette, and marijuana use

BACKGROUND

Neighborhood factors reported subjectively by residents and measured objectively at the census tract are both associated with adolescent alcohol, tobacco (cigarette and electronic cigarette), and other drug (marijuana) (ATOD) use. Less clear is how these neighborhood factors are longitudinally associated with each substance. Equivocal findings may be due to lack of consideration of individual, peer, and family effect modifiers, which could help adolescents overcome exposure to stressful neighborhood environments.

METHODS

We used multivariate logistic regressions with interaction terms to test whether parental monitoring, resistance self-efficacy (RSE) and being around peers who use ATOD modified the association between four subjective and objective neighborhood measures and odds of using each substance measured one year later among 2539 high school students and college freshmen originally recruited from middle schools in Southern California.

RESULTS

Census tract-level disadvantage was not longitudinally associated with ATOD use. However, perceptions of higher neighborhood disorganization, less social cohesion, and more neighborhood problems with alcohol and drug use were associated with higher odds of ATOD use. Higher RSE and weaker affiliations with peers who use ATOD consistently buffered negative effects of neighborhood disorganization and neighborhood problems with alcohol and drugs on past year ATOD use.

CONCLUSIONS

Community-level programs that increase social cohesion among neighbors, neighborhood monitoring of deviant behaviors, and better policing of open drug selling may prevent ATOD use. Programs should also target RSE and minimize affiliations with peers who use ATOD, which could reduce the magnitude of the association with ATOD, even for adolescents living in the most at-risk neighborhoods.

An analysis of the process and results of manual geocode correction

Geocoding is the science and process of assigning geographical coordinates (i.e. latitude, longitude) to a postal address. The quality of the geocode can vary dramatically depending on several variables, including incorrect input address data, missing address components, and spelling mistakes. A dataset with a considerable number of geocoding inaccuracies can potentially result in an imprecise analysis and invalid conclusions. There has been little quantitative analysis of the amount of effort (i.e. time) to perform geocoding correction, and how such correction could improve geocode quality type. This study used a low-cost and easy to implement method to improve geocode quality type of an input database (i.e. addresses to be matched) through the processes of manual geocode intervention, and it assessed the amount of effort to manually correct inaccurate geocodes, reported the resulting match rate improvement between the original and the corrected geocodes, and documented the corresponding spatial shift by geocode quality type resulting from the corrections. Findings demonstrated that manual intervention of geocoding resulted in a 90% improvement of geocode quality type, took 42 hours to process, and the spatial shift ranged from 0.02 to 151,368 m. This study provides evidence to inform research teams considering the application of manual geocoding intervention that it is a low-cost and relatively easy process to execute.

Identifying and characterizing hepatitis C virus hotspots in Massachusetts: a spatial epidemiological approach

Background

Hepatitis C virus (HCV) infections have increased during the past decade but little is known about geographic clustering patterns.

Methods

We used a unique analytical approach, combining geographic information systems (GIS), spatial epidemiology, and statistical modeling to identify and characterize HCV hotspots, statistically significant clusters of census tracts with elevated HCV counts and rates. We compiled sociodemographic and HCV surveillance data (n = 99,780 cases) for Massachusetts census tracts (n = 1464) from 2002 to 2013. We used a five-step spatial epidemiological approach, calculating incremental spatial autocorrelations and Getis-Ord Gi* statistics to identify clusters. We conducted logistic regression analyses to determine factors associated with the HCV hotspots.

Results

We identified nine HCV clusters, with the largest in Boston, New Bedford/Fall River, Worcester, and Springfield (p < 0.05). In multivariable analyses, we found that HCV hotspots were independently and positively associated with the percent of the population that was Hispanic (adjusted odds ratio [AOR]: 1.07; 95% confidence interval [CI]: 1.04, 1.09) and the percent of households receiving food stamps (AOR: 1.83; 95% CI: 1.22, 2.74). HCV hotspots were independently and negatively associated with the percent of the population that were high school graduates or higher (AOR: 0.91; 95% CI: 0.89, 0.93) and the percent of the population in the “other” race/ethnicity category (AOR: 0.88; 95% CI: 0.85, 0.91).

Conclusion

We identified locations where HCV clusters were a concern, and where enhanced HCV prevention, treatment, and care can help combat the HCV epidemic in Massachusetts. GIS, spatial epidemiological and statistical analyses provided a rigorous approach to identify hotspot clusters of disease, which can inform public health policy and intervention targeting. Further studies that incorporate spatiotemporal cluster analyses, Bayesian spatial and geostatistical models, spatially weighted regression analyses, and assessment of associations between HCV clustering and the built environment are needed to expand upon our combined spatial epidemiological and statistical methods.

Misclassification errors from postal code-based geocoding to assign census geography in Nova Scotia, Canada

OBJECTIVES

Postal codes are often the only available geographic identifiers in many sources of health data in Canada. In order to conduct geographic analyses, postal codes are routinely geocoded to census geography to link to ecological data. Despite common use of this method, the extent of geographic misclassification errors is poorly understood. We estimated misclassification errors in the geocoding of postal codes to assign census geography in Nova Scotia, Canada.

METHODS

We examined differences between counts and match rates for postal-code geocoded and actual locations of buildings in Nova Scotia at two census administrative area levels: dissemination areas (DAs) and census subdivisions (CSDs). Actual locations were based on the data collected by the provincial government containing actual latitude/longitude of buildings. Variation in misclassification by rurality, using Statistics Canada's classification, was also assessed.

RESULTS

Outside two urban areas (Halifax Metro and Sydney) which had <10% differences in counts, many DAs had >30% differences. Match rates showed similar patterns, with the vast majority of non-urban DAs having <40% match rates. Even in major urban areas, 10% of DAs had large misclassification errors. Misclassification errors at the CSD level were still too great to estimate counts or rates without further area aggregation.

CONCLUSION

Routine use of postal code geocoding should be replaced with geocoding of location information using additional identifiers such as civic addresses or latitude and longitude. If data holders did this in-house before providing data to researchers, the accuracy and capacity of geographic analysis would be enhanced while protecting confidentiality.

Smartphone-assisted spatial data collection improves geographic information quality: pilot study using a birth records dataset

It is well known that the conventional, automated geocoding method based on self-reported residential addresses has many issues. We developed a smartphone-assisted aerial image-based method, which uses the Google Maps application programming interface as a spatial data collection tool during the birth registration process. In this pilot study, we have tested whether the smartphone-assisted method provides more accurate geographic information than the automated geocoding method in the scenario when both methods can get the address geocodes. We randomly selected 100 well-geocoded addresses among women who gave birth in Alachua county, Florida in 2012. We compared geocodes generated from three geocoding methods: i) the smartphone-assisted aerial image-based method; ii) the conventional, automated geocoding method; and iii) the global positioning system (GPS). We used the GPS data as the reference method. The automated geocoding method yielded positional errors larger than 100 m among 29.3% of addresses, while all addresses geocoded by the smartphoneassisted method had errors less than 100 m. The positional errors of the automated geocoding method were greater for apartment/condominiums compared with other dwellings and also for rural addresses compared with urban ones. We conclude that the smartphone-assisted method is a promising method for perspective spatial data collection by improving positional accuracy.

Geographic Discordance Between Patient Residence and Incident Location in Emergency Medical Services Responses

STUDY OBJECTIVE

The location of a patient's residence is often used for emergency medical services (EMS) system planning. Our objective is to evaluate the association between patient residence and emergency incident zip codes for 911 calls.

METHODS

We used data from the 2013 National Emergency Medical Services Information System (NEMSIS) Public-Release Research Dataset. We studied all 911 calls with a valid complaint by dispatch, identifying zip codes for both the residence and incident locations (n=12,376,784). The primary outcomes were geographic and distance discordances between patient residence and incident zip codes. We used a multivariate logistic regression model to determine geographic discordance between residence and incident zip codes by dispatch complaint, age, and sex. We also measured distances between locations with geospatial processing.

RESULTS

The overall proportion of geographic discordance for all 911 calls was 27.7% (95% confidence interval [CI] 27.7% to 27.8%) and the median distance discordance was 11.5 miles (95% CI 11.5 to 11.5 miles). Lower geographic discordance rates were found among patients aged 65 to 79 years (20.2%; 95% CI 20.1% to 20.2%) and 80 years and older (14.5%; 95% CI 14.5% to 14.6%). Motor vehicle crashes (63.5%; 95% CI 63.5% to 63.6%), industrial accidents (59.3%; 95% CI 58.0% to 60.6%), and mass casualty incidents (50.6%; 95% CI 49.6% to 51.5%) were more likely to occur outside a patient's residence zip code. Median network distance between home and incident zip centroid codes ranged from 8.6 to 23.5 miles.

CONCLUSION

In NEMSIS, there was geographic discordance between patient residence zip code and call location zip code in slightly more than one quarter of EMS responses records. The geographic discordance rates between residence and incident zip codes were associated with dispatch complaints and age. Although a patient's residence might be a valid proxy for incident location for elderly patients, this relationship holds less true for other age groups and among different complaints. Our findings have important implications for EMS system planning, resource allocation, and injury surveillance.

Improving Surveillance of Sexually Transmitted Diseases through Geocoded Morbidity Assignment

OBJECTIVES

We assessed the added value of using a geocoder to improve sexually transmitted disease (STD) surveillance data and decision support through redistribution of inaccurately assigned morbidity in Richmond, Virginia.

METHODS

Virginia initiated geocoding of STD data as a data quality tool in 2002. Geocoded output files were assessed and discordant proportions of reported gonorrhea and chlamydia morbidity were reassigned appropriately for the city of Richmond, Chesterfield County, and Henrico County (2002 to 2006). We used Chi-square analysis to compare assignment proportions and calculated crude odds ratios for 2006 data to estimate increased case reassignment likelihood.

RESULTS

From 2002 to 2006, 149,229 cases of gonorrhea and chlamydia were reported within the Commonwealth of Virginia. Of the reported morbidity, 81% of cases (n=120,875) were successfully geocoded; 7% (n=8,461) of geocoded addresses were reassigned. Approximately 76% (n=6,412) of all reassigned cases occurred within Richmond and Chesterfield and Henrico counties. In 2006, 84% (n=654) of reassigned cases in this tri-city/county area were initially reported as Richmond morbidity. Data quality improvements reduced Richmond's artificially inflated morbidity by 18% and increased Chesterfield and Henrico morbidity by 17% and 55%, respectively. Richmond morbidity was three times more likely to be reassigned than Chesterfield cases (odds ratio [OR] = 2.93, 95% confidence interval [CI] 2.21, 3.90), and two times more likely than Henrico cases (OR=2.12, 95% CI 1.63, 2.76). Richmond's number one national rank for STD rates was reduced beginning in 2002.

CONCLUSIONS

Declining rates of STDs were statistically associated with geocoded morbidity reassignments. Implementation of this data quality business process has improved epidemiologic analyses, prevention planning, and assessment of resource allocations. The reduction in Richmond's national STD rankings is indicative of the effect geocoding can have on surveillance data.

Spatial measurement errors in the field of spatial epidemiology

Background

Spatial epidemiology has been aided by advances in geographic information systems, remote sensing, global positioning systems and the development of new statistical methodologies specifically designed for such data. Given the growing popularity of these studies, we sought to review and analyze the types of spatial measurement errors commonly encountered during spatial epidemiological analysis of spatial data.

Methods

Google Scholar, Medline, and Scopus databases were searched using a broad set of terms for papers indexed by a term indicating location (space or geography or location or position) and measurement error (measurement error or measurement inaccuracy or misclassification or uncertainty): we reviewed all papers appearing before December 20, 2014. These papers and their citations were reviewed to identify the relevance to our review.

Results

We were able to define and classify spatial measurement errors into four groups: (1) pure spatial location measurement errors, including both non-instrumental errors (multiple addresses, geocoding errors, outcome aggregations, and covariate aggregation) and instrumental errors; (2) location-based outcome measurement error (purely outcome measurement errors and missing outcome measurements); (3) location-based covariate measurement errors (address proxies); and (4) Covariate-Outcome spatial misaligned measurement errors. We propose how these four classes of errors can be unified within an integrated theoretical model and possible solutions were discussed.

Conclusion

Spatial measurement errors are ubiquitous threat to the validity of spatial epidemiological studies. We propose a systematic framework for understanding the various mechanisms which generate spatial measurement errors and present practical examples of such errors.

Neighborhood Racial Composition and Perceptions of Racial Discrimination: Evidence From the Black Women's Health Study

Little is known about the effects of social context or “place” factors (e.g., characteristics of local populations) on African Americans' perceptions and experiences of racism. Using data from 42,445 U.S. black women collected during the 1997 follow-up wave of the Black Women's Health Study, we investigated the association between neighborhood racial composition (“percent black” at the block-group level in 2000 Census data) and perceptions of racial discrimination. Perceived racial discrimination was measured using self-reports of the frequency of discrimination in “everyday” settings (e.g., being treated as if you are dishonest) and “lifetime” occurrences of discrimination on the job, in housing, and by the police. There was a linear inverse relationship between neighborhood percent black and perceived discrimination, i.e., higher percent black was associated with lower levels of discrimination. Our results support the conclusions that, relative to contexts in which blacks are a small minority, more evenly-mixed (i.e., integrated) contexts result in lower levels of discrimination (contact hypothesis), and mostly black contexts evidence the lowest levels of discrimination (ethnic density hypothesis).

Change in alcohol outlet density and alcohol-related harm to population health (CHALICE): a comprehensive record-linked database study in Wales

Background

Excess alcohol consumption has serious adverse effects on health and results in violence-related harm.

Objective

This study investigated the impact of change in community alcohol availability on alcohol consumption and alcohol-related harms to health, assessing the effect of population migration and small-area deprivation.

Design

A natural experiment of change in alcohol outlet density between 2006 and 2011 measured at census Lower Layer Super Output Area level using observational record-linked data.

Setting

Wales, UK; population of 2.5 million aged ≥ 16 years.

Outcome measures

Alcohol consumption, alcohol-related hospital admissions, accident and emergency (A&E) department attendances from midnight to 06.00 and violent crime against the person.

Data sources

Licensing Act 2003 [Great Britain.Licensing Act 2003. London: The Stationery Office; 2003. URL:www.legislation.gov.uk/ukpga/2003/17/contents(accessed 8 June 2015)] data on alcohol outlets held by the 22 local authorities in Wales, alcohol consumption data from annual Welsh Health Surveys 2008–12, hospital admission data 2006–11 from the Patient Episode Database for Wales (PEDW) and A&E attendance data 2009–11 were anonymously record linked to the Welsh Demographic Service age–sex register within the Secure Anonymised Information Linkage Databank. A final data source was recorded crime 2008–11 from the four police forces in Wales.

Methods

Outlet density was estimated (1) as the number of outlets per capita for the 2006 static population and the per quarterly updated population to assess the impact of population migration and (2) using new methods of network analysis of distances between each household and alcohol outlets within 10 minutes of walking and driving. Alcohol availability was measured by three variables: (1) the previous quarterly value; (2) positive and negative change over the preceding five quarters; and (3) volatility, a measure of absolute quarterly changes during the preceding five quarters. Longitudinal statistical analysis used multilevel Poisson models of consumption and Geographically Weighted Regression (GWR) spatial models of binge drinking, Cox regression models of hospital admissions and A&E attendance and GWR models of violent crime against the person, each as a function of alcohol availability adjusting for confounding variables. The impact on health inequalities was investigated by stratifying models within quintiles of the Welsh Index of Multiple Deprivation.

Results

The main finding was that change in walking outlet density was associated with alcohol-related harms: consumption, hospital admissions and violent crime against the person each tracked the quarterly changes in outlet density. Alcohol-related A&E attendances were not clinically coded and the association was less conclusive. In general, social deprivation was strongly associated with the outcome measures but did not substantially modify the associations between the outcomes and alcohol availability. We found no evidence for an important effect of population migration.

Limitations

Limitations included the absence of any standardised methods of alcohol outlet data collation, processing and validation, and incomplete data on on-sales and off-sales. We were dependent on the quality of clinical coding and administrative records and could not identify alcohol-related attendances in the A&E data set.

Conclusion

This complex interdisciplinary study found that important alcohol-related harms were associated with change in alcohol outlet density. Future work recommendations include defining a research standard for recording outlet data and classification of outlet type, the methodological development of residence-based density measures and a health economic analysis of model-predicted harms.

Funding

The National Institute for Health Research Public Health Research programme. Additional technical and computing support was provided by the Farr Institute at Swansea University, made possible by the following grant:

Centre for the Improvement of Population Health through E-records Research (CIPHER) and Farr Institute capital enhancement. CIPHER and the Farr Institute are funded by Arthritis Research UK, the British Heart Foundation, Cancer Research UK, the Chief Scientist Office (Scottish Government Health Directorates), the Economic and Social Research Council, the Engineering and Physical Sciences Research Council, the Medical Research Council, the National Institute for Health Research, the National Institute for Social Care and Health Research (Welsh Government) and the Wellcome Trust (grant reference MR/K006525/1).

Potential selection bias associated with using geocoded birth records for epidemiologic research

PURPOSE

There is an increasing use of geocoded birth registry data in environmental epidemiology research. Ungeocoded records are routinely excluded.

METHODS

We used classification and regression tree analysis and logistic regression to investigate potential selection bias associated with this exclusion among all singleton Florida births in 2009 (n = 210,285).

RESULTS

The rate of unsuccessful geocoding was 11.5% (n = 24,171). This ranged between 0% and 100% across zip codes. Living in a rural zip code was the strongest predictor of being ungeocoded. Other predictors for geocoding status varied with urbanity status. In urban areas, maternal race (adjusted odds ratio [aOR] ranging between 1.08 for Hispanic and 1.18 for black compared to white), maternal age [aOR: 1.16 (1.10-1.23) for ages 20-34 compared to <20], maternal nativity [aOR: 1.20 (1.15-1.25) for non-US versus US born], delivery at a birth center [aOR: 1.72 (1.49-2.00) compared to hospital delivery], multiparity [aOR: 0.91 (0.88-0.94)], maternal smoking [aOR: 0.82 (0.76-0.88)], and having nonprivate insurance [aOR: 1.25 (1.20-1.30) for Medicaid versus private insurance] were significantly associated with being ungeocoded. In rural areas, births delivered at birth center [aOR: 2.91 (1.80-4.73)] or home [aOR: 1.94 (1.28-2.95)] had increased odds compared to hospital births. The characteristics predictive of being ungeocoded were also significantly associated with adverse birth outcomes such as low birth weight and preterm delivery, and the association for maternal age was different when ungeocoded births were included and excluded.

CONCLUSIONS

Geocoding status is not random. Women with certain exposure-outcome characteristics may be more likely to be ungeocoded and excluded, indicating potential selection bias.

Tamoxifen and Antidepressant Drug Interaction in a Cohort of 16,887 Breast Cancer Survivors

BACKGROUND

Controversy persists about whether certain antidepressants reduce tamoxifen's effectiveness on lowering breast cancer recurrence. We investigated whether taking tamoxifen and antidepressants (in particular, paroxetine) concomitantly is associated with an increased risk of recurrence or contralateral breast cancer.

METHODS

We examined 16 887 breast cancer survivors (TNM stages 0-II) diagnosed between 1996 and 2007 and treated with tamoxifen in two California health plans. Women were followed-up through December 31, 2009, for subsequent breast cancer. The main exposure was the percent of days of overlap when both tamoxifen and an antidepressant (paroxetine, fluoxetine, other selective serotonin reuptake inhibitors, tricyclics, and other classes) were used. Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated using multivariable Cox regression models with time-varying medication variables.

RESULTS

Of the 16 887 women, half (n = 8099) used antidepressants and 2946 women developed subsequent breast cancer during the 14-year study period. We did not find a statistically significant increased risk of subsequent breast cancer in women who concurrently used paroxetine and tamoxifen. For 25%, 50%, and 75% increases in percent overlap days between paroxetine and tamoxifen, hazard ratios were 1.06 (95% CI = 0.98 to 1.14, P = .09), 1.13 (95% CI = 0.98 to 1.30, P = .09), and 1.20 (95% CI = 0.97 to 1.49, P = .09), respectively, in the first year of tamoxifen treatment but were not statistically significant. Hazard ratios decreased to 0.94 (95% CI = 0.81 to 1.10, P = .46), 0.89 (95% CI = 0.66 to 1.20, P = .46), and 0.85 (95% CI = 0.54 to 1.32, P = .46) by the fifth year (all non-statistically significantly). Absolute subsequent breast cancer rates were similar among women who used paroxetine concomitantly with tamoxifen vs tamoxifen-only users. For the other antidepressants, we again found no such associations.

CONCLUSIONS

Using the comprehensive electronic health records of insured patients, we did not observe an increased risk of subsequent breast cancer in women who concurrently used tamoxifen and antidepressants, including paroxetine.

Survey methodology of the geographic research on wellbeing (GROW) study

BACKGROUND

Causal inferences from survey research on health would benefit from population-based prospective survey designs. Because of decreasing survey response rates and residential mobility, however, loss to follow-up is of concern. The purpose of this paper is to describe the methodology of the geographic research on wellbeing (GROW) study and the resulting sample of women, their children, and their neighborhoods.

RESULTS

GROW (2012-2013) was designed as a follow-up mail/telephone survey of postpartum women who completed the statewide-representative Maternal and Infant Health Assessment (MIHA) baseline survey (2003-2007) in California. GROW was completed in English or Spanish by mothers whose index child from MIHA were aged 4-10 years. Its research focus is on the role of neighborhood environments on behavioral risk factors for cancer. The survey was developed based on expert guidance and extensive pilot testing and includes in-depth information on women's and children's health and behaviors, socioeconomic and demographic factors, psychosocial characteristics, and neighborhood perceptions, linked to objective neighborhood characteristics. The sample size for GROW is 3016 women. Response rates were 33% of the eligible sample and 75% of the active sample (those able to be located). GROW appears to be highly representative of its target population and its respondents lived in similar types of neighborhoods compared with all California neighborhoods.

DISCUSSION

Surveyed 5-10 years after baseline, the GROW mixed-mode methodology produced a prospective, representative sample of women with young children in California, comparing both individual and residential characteristics. The methods have implications for the 40 states and New York City that participate in CDC's Pregnancy Risk Assessment Monitoring System, as well as other cross-sectional studies with participants' contact information. Several recommendations for conducting similar follow-up studies with minimal loss to follow-up are provided.

Serum Phosphorus Levels, Race, and Socioeconomic Status in Incident Hemodialysis Patients

OBJECTIVE

We sought to examine the relationship between race, socioeconomic status, and serum phosphorus levels in patients with end-stage renal disease incident to hemodialysis (HD) at a large, integrated health-care delivery system in Southern California.

DESIGN

Retrospective cohort study.

SUBJECTS

A total of 5,778 adult patients who initiated HD at our institution between January 1, 2007 and June 30, 2013.

MAIN OUTCOME MEASURES

Unadjusted and adjusted time-averaged serum phosphorus levels and actual phosphorus levels over time. Phosphorus levels were also analyzed by repeated measures as a continuous measure and by phosphorus category. Baseline patient covariates included age, self-reported race, gender, cause of end-stage renal disease, and Charlson comorbidity index scores. Education and income level were estimated using geocoded data.

RESULTS

A total of 68,372 phosphorus levels were available for 4,862 patients. Estimated annual family income fell below $40,001 in 66.1% of African Americans (AAs) and 62.7% of Hispanics compared with 43.5% of Asians and 43.7% of whites, P < .0001. Educational level fell into the highest category for whites (70.8%) compared with AA (44.8%) or Hispanic (30.5%) patients, P < .0001. Adjusted time-averaged phosphorus levels were lower among Hispanics (4.33 mg/dL, 95% confidence interval [CI] 4.27-4.40) compared with Asian (4.54 mg/dL, 95% CI 4.45-4.64, P < .001) and white patients (4.48 mg/dL, 95% CI 4.43-4.54, P < .001) but similar to AA patients. Asian patients experienced a significant increase in phosphorus levels over time (0.11 mg/dL per year, P < .0001). There were no significant effects of race, time, or race by time interactions in the unadjusted and adjusted categorical analyses of phosphorus levels.

CONCLUSIONS

Our findings suggest that serum phosphorus levels are similar among HD patients, irrespective of race or socioeconomic status.

Tools for address georeferencing - limitations and opportunities every public health professional should be aware of

Various address georeferencing (AG) tools are currently available. But little is known about the quality of each tool. Using data from the EPIPorto cohort we compared the most commonly used AG tools in terms of positional error (PE) and subjects' misclassification according to census tract socioeconomic status (SES), a widely used variable in epidemiologic studies. Participants of the EPIPorto cohort (n = 2427) were georeferenced using Geographical Information Systems (GIS) and Google Earth (GE). One hundred were randomly selected and georeferenced using three additional tools: 1) cadastral maps (gold-standard); 2) Global Positioning Systems (GPS) and 3) Google Earth, single and in a batch. Mean PE and the proportion of misclassified individuals were compared. Google Earth showed lower PE than GIS, but 10% of the addresses were imprecisely positioned. Thirty-eight, 27, 16 and 14% of the participants were located in the wrong census tract by GIS, GPS, GE (batch) and GE (single), respectively (p<0.001). Misclassification according to SES was less frequent but still non-negligible −14.4, 8.1, 4.2 and 2% (p<0.001). The quality of georeferencing differed substantially between AG tools. GE seems to be the best tool, but only if prudently used. Epidemiologic studies using spatial data should start including information on the quality and accuracy of their georeferencing tools and spatial datasets.

Geocoding large population-level administrative datasets at highly resolved spatial scales

Using geographic information systems to link administrative databases with demographic, social, and environmental data allows researchers to use spatial approaches to explore relationships between exposures and health. Traditionally, spatial analysis in public health has focused on the county, zip code, or tract level because of limitations to geocoding at highly resolved scales. Using 2005 birth and death data from North Carolina, we examine our ability to geocode population-level datasets at three spatial resolutions - zip code, street, and parcel. We achieve high geocoding rates at all three resolutions, with statewide street geocoding rates of 88.0% for births and 93.2% for deaths. We observe differences in geocoding rates across demographics and health outcomes, with lower geocoding rates in disadvantaged populations and the most dramatic differences occurring across the urban-rural spectrum. Our results suggest highly resolved spatial data architectures for population-level datasets are viable through geocoding individual street addresses. We recommend routinely geocoding administrative datasets to the highest spatial resolution feasible, allowing public health researchers to choose the spatial resolution used in analysis based on an understanding of the spatial dimensions of the health outcomes and exposures being investigated. Such research, however, must acknowledge how disparate geocoding success across subpopulations may affect findings.

Predictors of peritonitis, hospital days, and technique survival for peritoneal dialysis patients in a managed care setting

INTRODUCTION

Many clinicians perceive that peritoneal dialysis (PD) should be reserved for younger, healthier, more affluent patients. Our aim was to examine outcomes for PD patients in a managed care setting and to identify predictors of adverse outcomes.

METHODS

We identified all patients who initiated PD at our institution between 1 January 2001 and 31 December 2010. Predictor variables studied included age, sex, race, PD modality, cause of end-stage renal disease (ESRD), dialysis vintage, Charlson comorbidity index (CCI) score, education, and income level. Poisson models were used to determine the relative risk (RR) of peritonitis and the number of hospital days per patient-year. The log-rank test was used to compare technique survival by patient strata.

RESULTS

Among the 1378 patients who met the inclusion criteria, only female sex [RR: 0.85; 95% confidence interval (CI): 0.74 to 0.98; p = 0.02] and higher education (RR: 0.77; 95% CI: 0.60 to 0.98; p = 0.04) were associated with peritonitis. For hospital days, dialysis vintage (RR: 1.11; 95% CI: 1.04 to 1.18; p = 0.002), CCI score (RR: 1.06; 95% CI: 1.02 to 1.20; p = 0.002), and cause of ESRD (RR for glomerulonephritis: 0.59; 95% CI: 0.43 to 0.80; p = 0.0006; and RR for hypertension: 0.69; 95% CI: 0.55 to 0.88; p = 0.002) were associated with 1 extra hospital day per patient-year. The 2-year technique survival was 61% for patients who experienced at least 1 episode of peritonitis and 72% for those experiencing no peritonitis (p = 0.0001). Baseline patient age, primary cause of ESRD, and PD modality were the only other variables associated with technique survival in the study.

CONCLUSIONS

Neither race nor socio-economic status predicted technique survival or hospital days in our study. Female sex and higher education were the only two variables studied that had an association with peritonitis.

Gender differences in associations of neighbourhood disadvantage with alcohol's harms to others: a cross-sectional study from the USA

INTRODUCTION AND AIMS

To examine whether alcohol's harms to others are more prevalent in socioeconomically disadvantaged neighbourhoods and whether men or women are at differential risk in these neighbourhoods.

DESIGN AND METHODS

Cross-sectional survey data from 2000 and 2005 National Alcohol Surveys were linked to geo-referenced indicators of neighbourhood disadvantage from the US 2000 Decennial Census. The pooled sample included 10,121 adults (54% female; average age 44.4 years; 69% White; 13% African-American; 13% Hispanic). A dichotomous indicator denoted neighbourhoods based on the top quartile on a five-item measure of disadvantage (alpha = 0.90). We examined past-year family problems due to someone else's drinking (marriage difficulties and/or financial trouble) and victimisation by someone who had been drinking (having property vandalised and/or being pushed, hit or assaulted).

RESULTS

During the prior 12 months, 6% of women and 3% of men experienced family problems from someone else's drinking, and 4% of women and 7% of men reported being victimised by drinkers. Multivariate logistic regression models adjusting for individual-level socioeconomic status and other demographic characteristics showed the relationship between neighbourhood disadvantage and harms from someone else's drinking was moderated by gender, with significantly higher odds of family problems in disadvantaged neighbourhoods for men but not for women, as well as significantly higher odds of crime victimisation in disadvantaged neighbourhoods for women but not men.

DISCUSSION AND CONCLUSIONS

Experiences of harms from someone else's drinking in disadvantaged neighbourhoods vary for men and women. Targeted intervention strategies are needed to reduce alcohol's harm to others.

Ensuring Confidentiality of Geocoded Health Data: Assessing Geographic Masking Strategies for Individual-Level Data

Public health datasets increasingly use geographic identifiers such as an individual's address. Geocoding these addresses often provides new insights since it becomes possible to examine spatial patterns and associations. Address information is typically considered confidential and is therefore not released or shared with others. Publishing maps with the locations of individuals, however, may also breach confidentiality since addresses and associated identities can be discovered through reverse geocoding. One commonly used technique to protect confidentiality when releasing individual-level geocoded data is geographic masking. This typically consists of applying a certain amount of random perturbation in a systematic manner to reduce the risk of reidentification. A number of geographic masking techniques have been developed as well as methods to quantity the risk of reidentification associated with a particular masking method. This paper presents a review of the current state-of-the-art in geographic masking, summarizing the various methods and their strengths and weaknesses. Despite recent progress, no universally accepted or endorsed geographic masking technique has emerged. Researchers on the other hand are publishing maps using geographic masking of confidential locations. Any researcher publishing such maps is advised to become familiar with the different masking techniques available and their associated reidentification risks.

An evaluation framework for comparing geocoding systems

BACKGROUND

Geocoding, the process of converting textual information describing a location into one or more digital geographic representations, is a routine task performed at large organizations and government agencies across the globe. In a health context, this task is often a fundamental first step performed prior to all operations that take place in a spatially-based health study. As such, the quality of the geocoding system used within these agencies is of paramount concern to the agency (the producer) and researchers or policy-makers who wish to use these data (consumers). However, geocoding systems are continually evolving with new products coming on the market continuously. Agencies must develop and use criteria across a number axes when faced with decisions about building, buying, or maintaining any particular geocoding systems. To date, published criteria have focused on one or more aspects of geocode quality without taking a holistic view of a geocoding system's role within a large organization. The primary purpose of this study is to develop and test an evaluation framework to assist a large organization in determining which geocoding systems will meet its operational needs.

METHODS

A geocoding platform evaluation framework is derived through an examination of prior literature on geocoding accuracy. The framework developed extends commonly used geocoding metrics to take into account the specific concerns of large organizations for which geocoding is a fundamental operational capability tightly-knit into its core mission of processing health data records. A case study is performed to evaluate the strengths and weaknesses of five geocoding platforms currently available in the Australian geospatial marketplace.

RESULTS

The evaluation framework developed in this research is proven successful in differentiating between key capabilities of geocoding systems that are important in the context of a large organization with significant investments in geocoding resources. Results from the proposed methodology highlight important differences across all axes of geocoding system comparisons including spatial data output accuracy, reference data coverage, system flexibility, the potential for tight integration, and the need for specialized staff and/or development time and funding. Such results can empower decisions-makers within large organizations as they make decisions and investments in geocoding systems.

Neighborhood socioeconomic status and substance use by U.S. adults

BACKGROUND

This study examined relationships of extremes in neighborhood socioeconomic status with use of tobacco, alcohol, marijuana and other drugs. Hypotheses were (1) residence in disadvantaged neighborhoods would be positively associated with stress-related and higher-risk substance use patterns (e.g., drug use), and (2) residence in affluent neighborhoods would be positively associated with "healthy" substance use (e.g., drinking within recommended guidelines) and negatively associated with substance use patterns incompatible with a culture of health. Age was examined as a potential moderator.

METHODS

Data were from nationally-representative samples of U.S. adults (N=14,531) from the 2000 and 2005 National Alcohol Surveys linked with indicators of neighborhood SES from the 2000 U.S. Decennial Census. Analyses included gender-stratified multivariate logistic regression using weights to adjust for sampling and non-response.

RESULTS

As hypothesized, compared to middle-class neighborhoods, residence in disadvantaged neighborhoods was associated with higher odds of both men's and women's tobacco use and with women's other drug use. Residence in affluent neighborhoods was associated with lower odds of men's tobacco use and women's marijuana use. The association of neighborhood SES with men's tobacco use was modified by age, with the highest odds of daily tobacco use evident for all men in disadvantaged neighborhoods, as well as for younger men in middle-class neighborhoods. There were no significant associations of either alcohol outcome with neighborhood SES.

CONCLUSIONS

Increased risk of substance use for younger residents in both disadvantaged and middle-class neighborhoods and for older residents in disadvantaged neighborhoods suggest a need for targeted prevention interventions.

Modeling travel impedance to medical care for children with birth defects using Geographic Information Systems

BACKGROUND

Children with birth defects may face significant geographic barriers accessing medical care and specialized services. Using a Geographic Information Systems-based approach, one-way travel time and distance to access medical care for children born with spina bifida was estimated.

METHODS

Using 2007 road information from the Florida Department of Transportation, we built a topological network of Florida roads. Live-born Florida infants with spina bifida during 1998 to 2007 were identified by the Florida Birth Defects Registry and linked to hospital discharge records. Maternal residence at delivery and hospitalization locations were identified during the first year of life.

RESULTS

Of 668 infants with spina bifida, 8.1% (n = 54) could not be linked to inpatient data, resulting in 614 infants. Of those 614 infants, 99.7% (n = 612) of the maternal residential addresses at delivery were successfully geocoded. Infants with spina bifida living in rural areas in Florida experienced travel times almost twice as high compared with those living in urban areas. When aggregated at county levels, one-way network travel times exhibited statistically significant spatial autocorrelation, indicating that families living in some clusters of counties experienced substantially greater travel times compared with families living in other areas of Florida.

CONCLUSION

This analysis demonstrates the usefulness of linking birth defects registry and hospital discharge data to examine geographic differences in access to medical care. Geographic Information Systems methods are important in evaluating accessibility and geographic barriers to care and could be used among children with special health care needs, including children with birth defects.

Field validation of secondary data sources: a novel measure of representativity applied to a Canadian food outlet database

Background

Validation studies of secondary datasets used to characterize neighborhood food businesses generally evaluate how accurately the database represents the true situation on the ground. Depending on the research objectives, the characterization of the business environment may tolerate some inaccuracies (e.g. minor imprecisions in location or errors in business names). Furthermore, if the number of false negatives (FNs) and false positives (FPs) is balanced within a given area, one could argue that the database still provides a “fair” representation of existing resources in this area. Yet, traditional validation measures do not relax matching criteria, and treat FNs and FPs independently. Through the field validation of food businesses found in a Canadian database, this paper proposes alternative criteria for validity.

Methods

Field validation of the 2010 Enhanced Points of Interest (EPOI) database (DMTI Spatial®) was performed in 2011 in 12 census tracts (CTs) in Montreal, Canada. Some 410 food outlets were extracted from the database and 484 were observed in the field. First, traditional measures of sensitivity and positive predictive value (PPV) accounting for every single mismatch between the field and the database were computed. Second, relaxed measures of sensitivity and PPV that tolerate mismatches in business names or slight imprecisions in location were assessed. A novel measure of representativity that further allows for compensation between FNs and FPs within the same business category and area was proposed. Representativity was computed at CT level as ((TPs +|FPs-FNs|)/(TPs+FNs)), with TPs meaning true positives, and |FPs-FNs| being the absolute value of the difference between the number of FNs and the number of FPs within each outlet category.

Results

The EPOI database had a "moderate" capacity to detect an outlet present in the field (sensitivity: 54.5%) or to list only the outlets that actually existed in the field (PPV: 64.4%). Relaxed measures of sensitivity and PPV were respectively 65.5% and 77.3%. The representativity of the EPOI database was 77.7%.

Conclusions

The novel measure of representativity might serve as an alternative to traditional validity measures, and could be more appropriate in certain situations, depending on the nature and scale of the research question.

Neighborhood socioeconomic status in relation to preterm birth in a U.S. cohort of black women

This study examines the association between neighborhood socioeconomic status (SES) and preterm birth among U.S. black women. A composite variable for neighborhood SES, derived from 7 U.S. Census Bureau indicators, was assessed in relation to self-reported preterm birth (505 spontaneous and 452 medically indicated) among 6,390 women in the Black Women's Health Study who delivered singleton births during 1995-2003. The odds ratio (OR) for preterm birth, comparing the lowest (most deprived) to the highest (least deprived) quartiles of neighborhood SES, was 0.98 (95% CI, 0.80, 1.20) after adjustment for individual-level characteristics. Low neighborhood SES was not associated with spontaneous or medically indicated preterm birth overall or within strata of maternal age, education, or geographic region. The only significant finding was higher odds of medically indicated preterm birth associated with low neighborhood SES among unmarried women. Low neighborhood SES was not materially associated with preterm birth in this study of U.S. Black women.

Childhood obesity is associated with increased risk of most lower extremity fractures

BACKGROUND

A number of studies have found an increased risk of lower extremity injuries in obese patients. Most studies, however, are unable to provide stable population-based estimates based on the degree of obesity and few assess the risk pertaining to more detailed fracture location in the lower extremities.

QUESTIONS/PURPOSES

We therefore investigated the relationship between obesity and lower extremity fractures in different age and fracture locations in a stable population.

METHODS

This is a population-based, cross-sectional study from the electronic medical records of 913,178 patients aged 2 to 19 years. The body mass index (BMI) for each patient in the cohort was used to stratify patients into five weight classes (underweight, normal weight, overweight, moderate obesity, and extreme obesity) based on BMI for age. Records were assessed for the occurrence of lower extremity fractures for each cohort member. The associations among the five weight classes and specific lower extremity fractures were estimated using multiple logistic regression models and expressed with odds ratios (ORs) and 95% confidence intervals (CIs) using multivariate analysis to adjust for patient demographic variables.

RESULTS

Overweight, moderately obese, and extremely obese patients all had an increased OR of fractures of the foot (OR, 1.14, 1.23, and 1.42, respectively, with 95% CI, 1.04-1.24, 1.12-1.35, and 1.26-1.61, respectively) along with the ankle, knee, and leg (OR, 1.27, 1.28, and 1.51, respectively, with 95% CI, 1.16-1.39, 1.15-1.42, and 1.33-1.72, respectively). The association was strongest in the 6- to 11-year-old age group. We found no association between increasing BMI and increased risk of fractures of the femur and hip.

CONCLUSIONS

Increasing BMI is associated with increased odds of foot, ankle, leg, and knee fractures in children.

LEVEL OF EVIDENCE

Level III, prognostic study. See Guidelines for Authors for a complete description of levels of evidence.

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.

Living in proximity of a bar and risky alcohol behaviours: a longitudinal study

AIMS

We examined whether distance from home to the nearest bar, i.e. alcohol outlet permitting consumption on the premises, is associated with risky alcohol behaviours.

DESIGN

Cross-sectional and longitudinal study.

SETTING AND PARTICIPANTS

The cross-sectional data consisted of 78 858 and the longitudinal data of 54 778 Finnish Public Sector Study participants between 2000 and 2009 [mean follow-up 6.8 years (SD = 2.0)].

MEASUREMENTS

Distances from home to the nearest bar were calculated using Global Positioning System coordinates. The outcome variables were heavy alcohol use (drinking above the weekly guidelines) and extreme drinking occasions (passing out because of alcohol use). We used binomial logistic regression in cross-sectional analyses and in longitudinal mixed effects (between-individual) analyses. Conditional logistic regression was used in longitudinal fixed effects (within-individual) analyses.

FINDINGS

Cross-sectionally, the likelihood of an extreme drinking occasion and heavy use was higher among those who resided <1 versus ≥ 1 km from a bar. Longitudinally, between individuals, a decrease from >1 km to ≤1 km in distance was weakly associated with an extreme drinking occasion [odds ratio (OR) 1.18, 95% confidence interval (CI) 0.98-1.41] and heavy use (1.12, 95% CI 0.97-1.29). Within-individual, the OR for becoming a heavy user was 1.17 (95% CI 1.02-1.34), per 1 km decrease in log-transformed continuous distance, the corresponding OR for an extreme drinking occasion was 1.03 (95% CI 0.89-1.18).

CONCLUSIONS

Moving place of residence close to, or far from, a bar appears to be associated with a small corresponding increase or decrease in risky alcohol behaviour.

Neighborhood disadvantage and adult alcohol outcomes: differential risk by race and gender

OBJECTIVE

We examined whether relationships of neighborhood disadvantage with drinker status, heavy drinking, alcohol-related consequences, and dependence differed by race and/or gender. We hypothesized that neighborhood disadvantage would be negatively associated with drinker status but positively associated with heavy and problem drinking, with more pronounced relationships among African American and Hispanic men than other groups.

METHOD

Data consisted of nationally representative, randomly selected, cross-sectional samples of White, African American, and Hispanic adults (N = 13,864, of which 52% were female; with 7,493 drinkers, of which 48% were female) from the 2000 and 2005 National Alcohol Surveys merged with 2000 Census data. Analyses included logistic and linear regression using weights to adjust for sampling and nonresponse.

RESULTS

Hypotheses were partly supported. Bivariate relationships were in the expected direction. Multivariate main effect models showed that neighborhood disadvantage was significantly associated with increased abstinence and marginally associated with increased negative consequences experienced by drinkers, but race/ethnicity and gender modified these associations. Disadvantage was significantly associated with increased abstinence for all groups except African American and Hispanic men. Among drinkers, disadvantage was significantly negatively associated with heavy drinking by Whites but significantly positively associated with heavy drinking by African Americans. Disadvantage also was associated with elevated alcohol-related consequences for White women and African American men.

CONCLUSIONS

The findings have implications for the development of targeted interventions to reduce the unequal impacts of neighborhood disadvantage on alcohol outcomes. Future research should examine the contribution of multiple types of disadvantage to heavy drinking and alcohol problems.

Socioeconomic characteristics of residential areas and risk of death: is variation in spatial units for analysis a source of heterogeneity in observed associations?

OBJECTIVES

Evidence on the association between the adverse socioeconomic characteristics of residential area and mortality is mixed. We examined whether the choice of spatial unit is critical in detecting this association.

DESIGN

Register-linkage study.

SETTING

Data were from the Finnish Public Sector study's register cohort.

PARTICIPANTS

The place of residence of 146 600 cohort participants was linked to map grids and administrative areas, and they were followed up for mortality from 2000 to 2011. Residential area socioeconomic deprivation and household crowding were aggregated into five alternative areas based on map grids (250×250 m, 1×1 km and 10×10 km squares), and administrative borders (zip-code area and town).

PRIMARY AND SECONDARY OUTCOME MEASURES

All-cause mortality.

RESULTS

For the 250×250 m area, mortality risk increased with increasing socioeconomic deprivation (HR for top vs bottom quintile 1.36, 95% CI 1.21 to 1.52). This association was either weaker or missing when broader spatial units were used. For household crowding, excess mortality was observed across all spatial units, the HRs ranging from 1.14 (95% CI 1.03 to 1.25) for zip code, and 1.21 (95% CI 1.11 to 1.31) for 250×250 m areas to 1.28 (95% CI 1.10 to 1.50) for 10×10 km areas.

CONCLUSIONS

Variation in spatial units for analysis is a source of heterogeneity in observed associations between residential area characteristics and risk of death.

Individual and neighborhood socioeconomic status in relation to breast cancer incidence in African-American women

Socioeconomic status (SES) for both individuals and neighborhoods has been positively associated with incidence of breast cancer, although not consistently. The authors conducted an assessment of these factors among African-American women, based on data from the Black Women's Health Study, a prospective cohort study of 59,000 African-American women from all regions of the United States. Individual SES was defined as the participant's self-reported level of education, and neighborhood SES was measured by a score based on census block group data for 6 indicators of income and education. Analyses included 1,343 incident breast cancer cases identified during follow-up from 1995 through 2009. In age-adjusted analyses, SES for both individuals and neighborhoods was associated with an increased incidence of estrogen receptor-positive breast cancer. The associations were attenuated by control for parity and age at first birth, and there was no association after further control for other breast cancer risk factors. These findings suggest that the observed associations of breast cancer with SES may be largely mediated by reproductive factors that are associated with both estrogen receptor-positive breast cancer and SES.

Effects of georeferencing effort on mapping monkeypox case distributions and transmission risk

BACKGROUND

Maps of disease occurrences and GIS-based models of disease transmission risk are increasingly common, and both rely on georeferenced diseases data. Automated methods for georeferencing disease data have been widely studied for developed countries with rich sources of geographic referenced data. However, the transferability of these methods to countries without comparable geographic reference data, particularly when working with historical disease data, has not been as widely studied. Historically, precise geographic information about where individual cases occur has been collected and stored verbally, identifying specific locations using place names. Georeferencing historic data is challenging however, because it is difficult to find appropriate geographic reference data to match the place names to. Here, we assess the degree of care and research invested in converting textual descriptions of disease occurrence locations to numerical grid coordinates (latitude and longitude). Specifically, we develop three datasets from the same, original monkeypox disease occurrence data, with varying levels of care and effort: the first based on an automated web-service, the second improving on the first by reference to additional maps and digital gazetteers, and the third improving still more based on extensive consultation of legacy surveillance records that provided considerable additional information about each case. To illustrate the implications of these seemingly subtle improvements in data quality, we develop ecological niche models and predictive maps of monkeypox transmission risk based on each of the three occurrence data sets.

RESULTS

We found macrogeographic variations in ecological niche models depending on the type of georeferencing method used. Less-careful georeferencing identified much smaller areas as having potential for monkeypox transmission in the Sahel region, as well as around the rim of the Congo Basin. These results have implications for mapping efforts, as each higher level of georeferencing precision required considerably greater time investment.

CONCLUSIONS

The importance of careful georeferencing cannot be overlooked, despite it being a time- and labor-intensive process. Investment in archival storage of primary disease-occurrence data is merited, and improved digital gazetteers are needed to support public health mapping activities, particularly in developing countries, where maps and geographic information may be sparse.

Using new satellite based exposure methods to study the association between pregnancy PM₂.₅ exposure, premature birth and birth weight in Massachusetts

BACKGROUND

Adverse birth outcomes such as low birth weight and premature birth have been previously linked with exposure to ambient air pollution. Most studies relied on a limited number of monitors in the region of interest, which can introduce exposure error or restrict the analysis to persons living near a monitor, which reduces sample size and generalizability and may create selection bias.

METHODS

We evaluated the relationship between premature birth and birth weight with exposure to ambient particulate matter (PM₂.₅) levels during pregnancy in Massachusetts for a 9-year period (2000-2008). Building on a novel method we developed for predicting daily PM₂.₅ at the spatial resolution of a 10x10 km grid across New-England, we estimated the average exposure during 30 and 90 days prior to birth as well as the full pregnancy period for each mother. We used linear and logistic mixed models to estimate the association between PM₂.₅ exposure and birth weight (among full term births) and PM₂.₅ exposure and preterm birth adjusting for infant sex, maternal age, maternal race, mean income, maternal education level, prenatal care, gestational age, maternal smoking, percent of open space near mothers residence, average traffic density and mothers health.

RESULTS

Birth weight was negatively associated with PM₂.₅ across all tested periods. For example, a 10 μg/m³ increase of PM₂.₅ exposure during the entire pregnancy was significantly associated with a decrease of 13.80 g [95% confidence interval (CI) = -21.10, -6.05] in birth weight after controlling for other factors, including traffic exposure. The odds ratio for a premature birth was 1.06 (95% confidence interval (CI) = 1.01-1.13) for each 10 μg/m3 increase of PM₂.₅ exposure during the entire pregnancy period.

CONCLUSIONS

The presented study suggests that exposure to PM₂.₅ during the last month of pregnancy contributes to risks for lower birth weight and preterm birth in infants.

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.

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.