Matching the Genotype in Resolution: Innovative Ways of Phenotype Capture

Summary:

Social and spatial context are important determinants of morbidity and mortality. However, there is little clarity about the role of context for kidney disease specifically, particularly before the end stage. Meanwhile, research clarifying the clinical, cellular, molecular, and genetic causes of kidney disease is accelerating considerably. We postulate that without contextual information, even the most detailed biomedical information cannot fully capture the factors that ultimately drive the development and progression of kidney disease. The Nephrotic Syndrome Study Network is integrating detailed, state-of-the-art information on a social and spatial context to enable the exploration of the associations between the social environment and kidney disease. Here, we discuss the extant literature on social context and kidney disease, present information on sources of contextual information, and provide recommended further reading to facilitate future research on the contribution of the social context to kidney disease.

Social and spatial context are important determinants of morbidity and mortality. Indeed, it is well established that countries, geographic regions, cities, and neighborhoods all vary in health behaviors,¹ incidence and prevalence of disease,^2–7 and mortality.^7,8 Furthermore, the inequalities in context are thought to be an important driver of inequalities in morbidity and mortality.^6,9 Indeed, one of the Public Health Infrastructure objectives of Healthy People 2010 was to integrate geospatial information in all major health data systems¹⁰ to clarify the role of place. Furthermore, Healthy People 2020 now includes specific indicators of the social environment, acknowledging its important role in population health and social health inequalities.¹¹ Despite the growing evidence that indicates the importance of context for the major risk factors and determinants of kidney disease—cardiovascular disease,^3–6 diabetes,⁹ obesity,^12–14 and human immunodeficiency virus (HIV)¹⁵—there is still a dearth of empiric research on the role of social and spatial contexts in relation to kidney disease itself, particularly at pre–end stage renal disease (ESRD) stages.^16,17

Meanwhile, research clarifying the clinical, cellular, molecular, and genetic causes of kidney disease is accelerating considerably. Although each of these stratum of information improves our overall understanding of disease, in isolation, data silos can lead to naive conclusions. One of the key principles of the Nephrotic Syndrome Study Network (NEPTUNE) is to consider multiple data strata to generate a greater understanding of the kidney diseases associated with nephrotic syndrome. We would argue further that, without contextual information, even the most detailed biomedical information cannot fully capture the factors that ultimately drive the development and progression of kidney disease. Now, NEPTUNE is adding an additional layer of information to include social and spatial context (also referred to as social environment or neighborhood) (Fig. 1), to support future research addressing the potential association between the social environment and nephrotic syndrome–associated kidney diseases.^{2,4–6,9,18–24} As shown in Figure 1, we are integrating multiple types of social and spatial contextual information that represents both the markers and potential mechanisms linking place with kidney disease.

Figure 1.

Representation of the original data capture in NEPTUNE with integration of new contextual information. miRNA, microRNA; SNP, single-nucleotide polymorphism.

Here, we discuss the extant literature on social context and kidney disease, present information on different sources of contextual information, and provide resources for additional contextual information. The NEPTUNE study linked this detailed and state-of-the-art contextual information; the information we provide here, and the data that we currently are compiling, can be linked to any clinical or epidemiologic data set with geospatial information for its participants. Notably, with little additional data collection burden to the study participant (eg, provision of address, consent to link address information), this linkage results in a rich set of contextual information relevant to kidney disease.

GEOSPATIAL INFORMATION AND KIDNEY DISEASE

Although there is still a dearth of empiric evidence linking social context to kidney disease, there is some evidence that there is geographic variation in certain markers of kidney health, particularly ESRD-related measures.^17,25,26 Figure 2A shows the state-level variation in the age-, sex-, and race-adjusted incidence of ESRD per million population. This likely is owing to the geographic variation in socioeconomic and sociopolitical (generalized to social) environments that dictate the balance of resources and constraints relevant to overall health. Figure 2B shows the state-level percentages of children living at or below the poverty line; we show child poverty rather than overall poverty because this indicator of the social environment of the most vulnerable population may better capture the underlying resources that a state allocates to the quality of the overall social environment.²⁷ Note the variation in these two maps, in the overlap between ESRD incidence (that primarily is in adults²⁸) and child poverty, which suggests common social and spatial determinants. Therefore, in addition to documenting this geographic variation in kidney disease, we are moving toward clarifying the features of the social environment that drive this variation. It may be that the local environment, namely the residential neighborhood, is the link that explains the geographic variation in kidney disease. Clarifying the role of neighborhoods is critical to health because neighborhood features are neither random nor naturally occurring, meaning that they are amenable to policies and other interventions¹⁹ and may be an effective strategy for promoting population health.²⁹

Figure 2.

Map of spatial distribution of (A) ESRD incidence and (B) child poverty level. (A) ESRD incidence per million population, 2011. Data from the US Renal Data Service. (B) Percentage of children younger than age 15 living below the poverty level, 2010. Data from the US Population and Household Census.

Types of Geospatial Information

There is a wealth of geospatial information available from the federal government and we outline a portion of this information that may have particular relevance to kidney disease in Tables 1 through 3. Although we discuss the integration of federal data, there also are local data available to varying degrees. Because there has been a recent overview of the methods used for linking geospatial information to kidney disease data,¹⁶ we focus on the substantive issues regarding the importance of geospatial information.

Table 1.

Examples of Publicly Available Geographically Linked Information for Sociodemographic Context

						Suggested References
Concept	Suggested Variables	Data Source	Lowest Geographic Level Available	Years Available	Suggested Measures	Substantive Overview	Methods
SES	Median household income	Census	Census: block	Census: at least 1980 to present	Factor analysis–based multi-item z-score indices	30,31,87,88	36,44,60,61
	Median value of housing units	ACS	ACS: tract in 5-year aggregate; higher levels available annually	ACS: 2005 to present	City-level economic or educational spatial residential segregation
	Percentage of households with interest, dividend, or rental income	CPS	CPS: state; 12 largest metropolitan statistical areas	CPS: 1989 at CPS website; before 1989 from other public sources (NBER, ICPSR, IPUMS)	Neighborhood-level economic or educational spatial residential segregation
	Percentage of residents age ≥25 with a high school diploma
	Percentage of residents age ≥25 with a college degree
	Percentage of residents in executive, managerial, or professional specialty occupations
	Percentage of unemployment
Race/ethnicity	Percentage of a particular racial/ethnic group in a neighborhood or city	Census	Census: block	Census: any year, with categories changing over time	City-level racial/ethnic spatial segregation	59	60–63
		ACS	ACS: tract in 5-year aggregate; higher levels available annually	ACS: 2005 to present	Neighborhood-level racial/ethnic spatial segregation

NOTE. Area-level SES is not a proxy for individual-level SES but captures contextual information that is more than the sum of the individuals.^30,32 Research indicates the examination of affluence as well as deprivation.⁶

Abbreviations: Census, US Household and Population Census; CPS, US Current Population Survey (www.census.gov/cps); IPUMS, Integrated Public Use Micro Data Series (www.ipums.org); ICPSR, Inter-University Consortium for Political and Social Research (www.icpsr.umich.edu); NBER, National Bureau of Economic Research (www.nber.org).

Table 3.

Examples of Publicly Available Geographically Linked Information for Health Care Resources

						Suggested References
Concept	Possible Variables	Data Source	Lowest Geographic Level Available	Years Available	Possible Measures	Substantive Overview	Methods
Health facilities	Number per area or per population	AHRF	AHRF: county for some; State for most	AHRF: 1980 to present for many variables; 2005 to present for others	It will be informative to examine different types of hospitals (eg, community, long term) and clinics (eg, emergent care centers)	93–95	96
	Number of beds per area or population
	Admissions per population
Physicians	Number per area or population	AHRF	AHRF: county for some; state for most	AHRF: 1980 to present for many variables; 2005 to present for others	It will be informative to examine primary care separately from specialties and physicians separately from allied health personnel
Dialysis centers	Number per ESRD population	CMS	CMS: facility address	CMS: varies	With addresses of both the center and the patient, it is possible to use novel distance decay approaches	97,98	98,99
	Number of chairs per ESRD population	USRDS	USRDS: Zip code	USRDS: 1988 to present
	Mortality rate
Other health care resources	Health care expenditures; proportion covered by public and/or private health insurance; number per area and per population of pharmacies and other allied health centers	CMS	Census: block	CMS: varies to present	To capture the overall health care environment, it is appropriate to create a factor analysis–based z-score index of multiple variables, as described for socioeconomic status in Table 1	97
		Census	ACS: tract in 5-year aggregate; higher levels available annually	Census: varies to present
		ACS		ACS: 2005 to present

NOTE. There are numerous sources for geospatially linked health care information. However, most information is available at the state level. The association between health care resources and health outcomes may not be consistent. It may be that these resources reflect investment in the area to promote health, but it also may be that these resources are spatially located in areas where the need is greatest, in areas where residents have greater morbidity.

Abbreviations: AHRF, Area Health Resource File (www.arhf.hrsa.gov); Census, US Household and Population Census (www.census.gov/econ); CMS, Centers for Medicaid and Medicare Services (www.cms.gov); USRDS, US Renal Data System (www.usrds.org).

There are two common levels of geospatial information that can be linked to study participant health information: city level and neighborhood level. Generally, city-level information is captured along the boundaries of metropolitan statistical areas, which comprise one or more entire counties with a high degree of social and economic integration (except New England, where cities and towns are the basic geographic units). Linking individual-level (ie, participant-level) biomedical information to city-level contextual information may provide clues as to different policies that can improve outcomes. Policies that may differ across cities include those that are more social (eg, investments in education or the middleclass employment sector) or more directly health-related (eg, investments in primary care or dialysis centers).

One also can examine geospatial information in relation to kidney disease at the level of the residential neighborhood. Conceptually speaking, residential neighborhoods are considered to be areas within cities, often with some cohesive defining element, that element may include such definitions as a historical boundary, an important landmark, or the cluster of an ethnic group. Neighborhoods are thought to be related to health through local resources (eg, healthy food, safe spaces) and social norms (eg, exercise is common).

When measuring neighborhoods, researchers most often use governmental administrative boundaries because of their convenience and the amount of data available at these levels. Most commonly used are Census tracts, which are small, relatively permanent statistical subdivisions of a county with an optimum population size of 4000 people. In addition, Census block groups are used, which are divisions of the tract that contain approximately 600 to 3000 people. These Census-based measures are created and maintained specifically to track population. Zip codes, on the other hand, represent groups of approximately 30,000 addresses (Fig. 3) and are created and maintained for the efficient delivery of mail. Although these addresses within Zip codes generally are contiguous, a single Zip code often includes different neighborhoods and shows greater heterogeneity with regard to social factors.¹⁶ These administrative boundaries generally represent areas larger than the conceptualization of neighborhoods, particularly when using Zip codes. Although the Census does include information at the more granular block group level, there are other issues to consider, such as: research suggests that information at the tract and block group levels are moderately to highly correlated,³⁰ and examining rare diseases in small geospatial areas may increase the risk of re-identification even when using limited or de-identified data sets (discussed further by Rodriguez et al¹⁶). Confidentiality issues must be managed prospectively to align with consent and regulatory approvals to ensure adequate protections of patients.

Figure 3.

Census tracts and Zip codes for Detroit, MI, 2010. Reprinted from Data Driven Detroit, www.datadrivendetroit.org.

Linking individual data to neighborhood data can provide information about the quality and quantity of the social and health-related resources available and the condition of the overall social environment. Furthermore, as we discuss later, the unequal distribution of resources and the unequal social environments within cities may provide clues as to why some social groups show greater incidence, prevalence, and progression of kidney disease. Finally, in addition to these two levels, one also can examine the interaction between the two. For example, it may be that neighborhood-level resources are related to kidney disease in one city but not another, owing to city-level variation in resources.

There is considerable geospatial information available on national US (Tables 1–3) and local scales. We highlight three categories of geospatial information of particular relevance to kidney disease: sociodemographic context, the built environment, and health care resources. These categories were selected because they show the range of information available and are the basis of linkage activities ongoing in the NEPTUNE consortium.

SOCIODEMOGRAPHIC CONTEXT

Sociodemographic context generally refers to the socioeconomic status (SES) and demographic characteristics of an area. Although these characteristics are a composite of the individuals in the area, they capture more than that the sum of the individual residents³¹ and are not simply a proxy for individual-level characteristics.^31,32 The sociodemographic context in which a person lives is thought to be important for disease development and progression because it is a marker of the resources, constraints, and social norms of the area that may be important for disease. A growing body of research indicates the importance of place, particularly neighborhood, for kidney disease and its related comorbidities including hypertension, obesity, and diabetes.^2,18,33

The two characteristics in this category most commonly examined in relation to disease in epidemiologic studies are SES and racial/ethnic composition. Although the study of contextual factors and kidney disease is still in its early stages, some investigators have begun to examine neighborhood-level SES and race/ethnicity (particularly black race) and kidney disease. With regard to neighborhood-level SES and kidney disease, the results have been mixed.^34–40 This is likely owing to at least three reasons. First, most studies use a single measure of neighborhood SES, most often a marker of disadvantage, such as the proportion of the neighborhood in poverty. These single measures likely do not fully capture the contextual factors that are important for kidney disease. For this reason, some investigators have begun to explore the use of multi-item indices that capture multiple SES-related features.^34,36,37

Second, sociologic research suggests that deprivation is not the only important dimension important for chronic disease and that neighborhood affluence, particularly within the past 20 years, may be related more strongly to chronic diseases.^6,41 Indeed, in the studies of neighborhood-level SES and kidney disease, results consistently have shown that the broad continuum of neighborhood SES (marked by an SES index) is related inversely to kidney disease.^36,37,42

Third, conceptually speaking, it may not be that the SES of one’s neighborhood, in isolation, captures features that are salient for kidney disease. Rather than examining one’s neighborhood separate from the surrounding neighborhoods and the larger city overall, there is a need to examine residential segregation, which is the residential separation of different social groups (eg, wealthy and poor, black and white). Although the social science literature has documented the growing importance of socioeconomic segregation (eg, segregation by educational level),^43,44 there is only one known study of economic segregation and kidney disease. This research suggests that the clustering of counties with low and high SES is related to high and low ESRD incidence, respectively.⁴²

Racial/ethnic residential segregation is another type of segregation and is a process by which different racial/ethnic groups are sorted into neighborhoods of unequal quality through historical and current discriminatory policies and practices.^45–50 Racial/ethnic segregation is thought to be an important mechanism by which racial/ethnic inequalities in health are produced and maintained through differential access to economic opportunities and health-related resources and differential exposure to psychosocial stress and environmental hazards.^45,50–54

Although there are still only a handful of studies on the relationship between racial/ethnic segregation and kidney disease, research has suggested that living in neighborhoods with a greater proportion of black residents is associated with a lower likelihood of appearing on renal transplant waitlists, longer renal transplant times, and a greater risk of mortality on dialysis compared with living in neighborhoods with a lower proportion of black residents.^55–58 Beyond these ESRD outcomes, living in a neighborhood with a greater proportion of black residents also is associated with kidney disease comorbidities such as obesity and hypertension, independent of one’s own race.^3,13,14,57

Measures of residential segregation, whether economic or racial/ethnic, city or neighborhood level, have been improved substantially, particularly within the past 15 years. However, these improved measures generally have not been incorporated into health research. Most research on segregation and health still use aspatial measures such as the proportion of a racial group in a Census tract (for neighborhood-level segregation) or the index of dissimilarity (for city-level segregation). These measures are only a proxy for the underlying conceptualization of segregation (ie, clustering and separation of social groups) because they do not link the tract to neighboring tracts or the larger city (in the case of neighborhood-level measures), or the spatial separation of neighborhood clusters within a city (in the case of city-level measures), to capture the clustering of groups in space (e.g., their locations within the city and in relation to other neighborhoods) that likely mirror the spatial distribution of resources and constraints.⁵⁹ In other words, they are “aspatial” or without reference to their position in the space of surrounding tracts or the larger city. This is important to consider when examining neighborhood influences on kidney disease as a poor tract likely means something different when in a poor versus an affluent city. We currently are linking state-of-the-art spatial measures of neighborhood city segregation^44,60–63 to the NEPTUNE data set and will have these measures available for other kidney disease cohorts as well.

Data Sources

The primary source for geospatial information on sociodemographic context is the US Population and Household Census (Census) summary file 3 (through 2000) and the American Community Survey (ACS) (2005 to present) (Tables 1–3). Although there is detailed sociodemographic information on a subset of the population in the decadal Censuses before 2000, beginning in 2005, the US Census Bureau began to administer the ACS on a 3% subsample of the population annually. Census and ACS data can be obtained free of charge from the US Census Bureau (www.census.gov/econ) and the Integrated Public Use Micro Data Series of the Minnesota Population Center (www.ipums.org). It is important to note that the administrative boundaries (ie, Census tract, metropolitan statistical area) and definitions change from time to time, requiring the use of a standardized Census data set, called the Neighborhood Change Database (available for a fee from Geolytics, Inc, www.geolytics.com), if temporal comparisons are to be made.

THE BUILT ENVIRONMENT

The geospatial variation in kidney disease may be owing to variation in the built environment. The concept of the built environment refers to a constellation of human-made surrounding features that together are the setting for human activity. The built environment comprises discrete features such as buildings, sidewalks, and parks, to name only a few. Each of these features alone may have a relationship to health more broadly and to kidney disease more specifically. For example, the number and spatial distribution of covered bus stops in cities may affect the ease with which residents reach their health care appointments. However, the quantity, and particularly the quality, of these features in composite may have more measureable effects on health.

Although there is no known research linking the built environment and kidney disease, there is a growing amount of literature on the importance of the built environment to kidney disease comorbidities, particularly obesity. Indeed, there is a wealth of studies on walkability and healthy food resources, most often in relation to obesity, with many, but not all, finding an inverse association (see Papas et al,⁶⁴ Booth et al,⁶⁵ and Feng et al⁶⁶ for reviews), and others found that inequalities in these features may explain inequalities in obesity.⁶⁷ Research into obesity-related outcomes including diabetes and cardiovascular disease have been speculated,^18,68 but the empiric literature has yielded mixed results,^{12,21,22,66,68–71} which may be owing to incomplete measurement of the built environment.

Data Sources

Toward this conceptual framework of the built environment as a whole, researchers have developed indices of its features to examine the relation of the built environment to mental and physical health. Although public health researchers have focused more on access to healthy food and walkability features that often rely on surveys and patient reports, social scientists have developed more encompassing measures based on freely available information from Google Earth (Mountain View, California) (www.google.com/earth).^72,73 A set of geospatial boundaries and the audit routine (eg, audit all street faces, only left-side faces, only even blocks, and so forth) are determined by the researcher based on the specific research question and the research resources available. An important thing to consider when using these Google-based audit methods is that there is room for qualitative interpretation by the raters and, therefore, rater training and testing for consistency is critical.

Once the boundaries and routine are set and the raters have been trained and tested, the research team can begin the audit. In the most comprehensive versions of these indices,^72,73 nine categories of features are measured for both their quantity and quality: land use (which includes the type of businesses, such as grocery stores and pawn shops), buildings, sidewalks, pedestrian access, traffic design, amenities (eg, benches, garbage cans), esthetics and design, nature, and disorder (eg, vandalism, abandoned property). Within each of these categories are multiple measures that then are combined to create an overall score for the feature. Each score then can be modeled separately or together in relation to kidney disease.

HEALTH CARE RESOURCES

Another reason for geospatial variation in kidney disease is variation in the quantity and quality of health care resources. It may be that the number of placement of dialysis centers, tertiary care hospitals, and nephrologists in an area is associated with ESRD comorbidities and mortality. For example, dialysis survival is reported to be worse than expected in for-profit compared with not-for-profit dialysis centers,^58,74,75 although other investigators reported no association between for-profit status and outcomes.⁷⁶ There is a dearth of information on the associations of other types of health care resources and kidney disease. It also may be that the number and placement of primary care physicians and community-based hospitals is important for the early recognition and management of kidney disease risk factors that influence kidney disease incidence and outcomes. These health care resources, along with the number of people in an area covered by either public or private insurance and health professions training institutions (eg, medical schools) may indicate broader civic investments in and social norms about health overall.

Data Sources

There are three primary sources of data on health care resources. First, the Area Health Resource File (www.arhf.hrsa.gov) is a set of freely available health data resource products drawn from an extensive county-level database assembled annually from more than 50 sources by the US Health Resources and Services Administration. The Area Health Resource File products include geospatial information on the number of the following: (1) health professionals detailed by specialty (eg, internal medicine physician, nurse practitioner, dentist, and so forth); (2) health profession training detailed by type (eg, medical schools, nursing schools, and so forth); (3) health facilities detailed by type (eg, tertiary care center, hospice, and so forth); and (4) visits to health facilities (including length of stays). Information also is available on hospital expenditures.

A second source of health care information is the Census Bureau (www.census.gov/topics/health.html) through their administration of several data programs and the curation of data from other administrative programs: Census, ACS, Current Population Survey, Survey of Income and Program Participation, County Business Patterns, Nonemployer Statistics, Economic Census, Service Annual Survey, and Statistics of US Businesses. Although there is information on disability, fertility, and HIV/acquired immune deficiency syndrome, perhaps more relevant to health care resources, there is information on expenses and investments on health care institutions and health insurance coverage.

Finally, the Centers for Medicare and Medicaid Services (www.cms.gov) provide information about the location and quality of dialysis centers in publicly available files. These data are easily accessible through their Data Navigator system (www.dnav.cms.gov) and provides information linked to the county level. However, because they also include the exact addresses for the dialysis centers, researchers can link this information to the Census tract through the Census Bureau’s geocoder tool (www.geocoding.geo.census.gov/geocoder/). Because most tracts do not contain a dialysis center, this information is needed only if dialysis center information is integrated into more general information on neighborhood health care resources or one is examining specific cities. Otherwise, the city- and state-level data from the Centers for Medicare and Medicaid Services is appropriate.

SUMMARY AND OTHER CONSIDERATIONS

We live in a time of unprecedented access to social science data that can be linked to biomedical data to improve our perspective and understanding of human health and disease. The NEPTUNE study is a prime example of a rich data source from phenome to genome generated to define the molecular basis of disease and disease course. The inclusion of social contextual data will enhance our ability to explore additional factors that influence human health and therefore improve the precision of our conclusions and expand our interventions to optimize the health of individuals affected by nephrotic syndrome.

The major challenge to linking detailed geospatial information to clinical cohort studies is that, although there is little burden on participants, there is considerable effort required from the research team to access and manage the needed variables. Despite the effort required, there is substantial information to be gained on the role of social context in the development and progression of kidney disease. This contextual information is important because it captures the resources and constraints that ultimately drive numerous factors related to kidney disease including health behaviors, health care quality, and exposure to social stressors and environmental hazards.

There are methodologic implications for the use of contextual data in relation to kidney disease as well. When examining individuals linked to their context, it may be necessary to use multilevel models that account for the nonindependence of people within their geospatial areas and that also allow for the decomposition of area-level and individual-level factors in relation to kidney disease.³¹ It may be that smaller clinical cohort studies, particularly those of rare diseases, may not require these advanced analytic methods because there is no substantial clustering of patients within each area (eg, tract). In these cases, clustered robust standard errors may be sufficient to account for any nonindependence among patients.

Although gathering clinical, cellular, molecular, and genetic information will assist in the development of a personalized medical care plan, linking contextual information to clinical cohort data will allow for the examination of factors that drive the early development of kidney disease progression and provide for additional points of intervention. Notably, the contextual points of intervention not only may improve kidney health for the individual patient, but also may improve the overall health of the patient and of the population more broadly. Because kidney disease is linked tightly to numerous other conditions (eg, hypertension, cardiovascular disease, obesity, diabetes, and HIV) that are related to social context, targeting context for intervention for broader health along with targeting kidney disease clinical points, there may be greatly improved patient outcomes and overall population health.

Although we focused on linking geospatial information to individual patients, it also is informative to examine the link between composite spatial information (eg, city-level racial/ethnic segregation) and composite health outcomes (eg, racial/ethnic differences in mortality with ESRD). These are considered ecologic models and can be easier to use, require less data, and may preserve confidentiality when mapping rare diseases. However, with these models, inferences cannot be made about individuals who comprise these composite measures.

We also focused on residential area. However, on average, most Americans spend their waking time at work.⁷⁷ However, there is little information on the role of the occupational neighborhood on health more generally, and kidney disease more specifically. It may be that the built environment and health care resources of the occupational environment is an important component of a person’s overall social context.

We focused on social context within the US setting, however, kidney disease is clearly a global concern.^78–86 When examining social context in other countries, consideration must be given both to the geospatial definitions (eg, what is a neighborhood?) and the feature measurement (eg, what measure of SES is related to health? what neighborhood features mark underlying processes important to health?). Limitations in data availability often will determine both the geospatial definition and feature measurement, but collaboration with global partners to determine the best approaches to measurement is advised.

With the growing recognition that there are numerous factors beyond the individual patient that are important for kidney disease development and progression and that contextual place is important for health, there is need for research on the role of the social context in kidney disease. Indeed, recent reviews have indicated the paucity of information that is known about the social environment in relation to kidney disease, particularly at earlier stages.^16,17 Future research should focus on both the sociodemographic characteristics of the social environment and the potential mechanisms (eg, built environment and health care resources) while using the most appropriate measures informed by the social sciences. The linkage of geospatial measures within the NEPTUNE data set will be a model for additional kidney disease data sets across the United States.

Table 2.

Examples of Publicly Available Geographically Linked Information for the Built Environment

Concept	Suggested variables*
Land use	Undeveloped, residential, industrial, commercial, institutional, and public spaces
Buildings	Types and features of residential, commercial, and institutional buildings
Sidewalks	Presence, features, and quality of sidewalks
Pedestrian access	Features such as pedestrian access among neighborhood, public spaces, and commercial areas; includes information on barriers and access across major roadways and industrial areas
Traffic design	Type and quality of roadways, signage, blind spots such as driveways
Public amenities	Features such as bike lanes, water fountains, libraries, parks, and garbage cans
Commercial amenities	Food environment, which includes information such as types of food stores and restaurants and bars and other alcohol stores; other commercial outlets, which includes places such as banks and check-cashing services, gyms, dry cleaners, clothing stores, and gas stations
Aesthetics and design	Features on homes and buildings, home and building distance from roads, and enclosures
Disorder	Features such as the presence of graffiti and other types of vandalism, garbage, and abandoned property

NOTE. The data source for each of these variables was Google Earth (www.google.com/earth). Additional information on commercial amenities can be found in the US Economic Census (www.census.gov/econ). The lowest geographic level of data available for Google Earth is at the address point. To obtain area-level composite indices, one collects information within the street boundaries of the area (eg, Census tract). The lowest geographic level for the Economic Census is the county for most variables. Google Earth has only current information. The Economic Census has information every 5 years, with varying degrees of information with each census. Suggested measures are all factor analysis–based indices. For further reading on the substantive topic of the built environment, see Srinivasan et al⁹⁰ and Brownson et al.⁹¹ For further reading on the methodologic approach to measurement, see Hwang and Sampson,⁷² Bader,⁷³ Clarke et al,⁸⁹ and Bader et al.⁹²

^*For a complete set of variables, see Hwang and Sampson,⁷² Bader,⁷³ and Clarke et al.⁸⁹