The socio-spatial neighborhood estimation method: An approach to operationalizing the neighborhood concept
Highlights
► We review issues with previous health research using neighborhood analysis. ► We conceptualize and then operationalize neighborhoods using a new approach. ► The socio-spatial neighborhood estimation method is explained and then evaluated. ► Validity evidence using a large sample and multi-level models support the method. ► We conclude that the method could improve inferences about neighborhoods and health.
Introduction
Research aimed at discovering how the social and physical environment affects health and well-being has increased rapidly during the last decade. A significant proportion of that inquiry focuses on the effects of so-called “neighborhoods.” Recognition of concomitant methodological difficulties in neighborhoods and health research, however, is growing. Further, such methodological difficulties ultimately impact the quality of inferences with regard to health outcomes. One of the most compelling challenges to the quality of inferences from this literature is the definition of the neighborhood as an ecological unit of analysis (Riva et al., 2007). In this paper, we argue that central among those challenges is the development of a methodology to create neighborhood boundaries that define areas in a conceptually valid manner (as opposed to census tracts, for example). Such a method would also help generate more reliable measures of social ecologies, which can then be linked to a host of well-being measures. Moreover, the challenge includes the need to balance the limitations inherent in any such method with the difficulty of use in research. The ultimate goal in addressing these issues in social epidemiology is to improve inferences about the relationship between a very important type of place in human experience—the neighborhood—and various dimensions of health and well-being.
Accordingly, colleagues involved in the Texas City Stress and Health Study developed a new methodology—the socio-spatial neighborhood estimation method (SNEM). SNEM was designed for creating conceptually informed neighborhood boundaries. This method became the basis for data collection and, eventually, multi-level modeling of ecological and individual variables. The SNEM approach employs principles of neighborhood boundary definition along with remote (aerial or satellite) imagery to draw boundaries and create neighborhoods for research purposes. Although innovative, the approach is relatively easy to employ by any researcher with access to a basic geographical information system in which to view aerial images of urban space.
In this report, we provide a rationale for this approach and explain its potential significance for the small area effects (on health) literature. We then describe and illustrate the method. To test the value of the approach, we report a two-step “ecometric” (Raudenbush and Sampson, 1999) comparison of ecological measures obtained using the SNEM areas versus the same measures obtained with census block groups and regular grid cells in the study area. Comparisons of intra-class correlation coefficients (ICCs) and establishment of criterion-related validity by way of multi-level models support the conclusion that the areas created with the SNEM approach are more neighborhood-like and provide a better basis for measurement and inferences about neighborhood effects than census boundaries or a regular grid matrix. We conclude with a discussion of the merits and limitations of the SNEM methodology for health research.
Section snippets
Background
The rise of social epidemiology and interest in neighborhoods and health was motivated by research on stress as well as the re-emergence of ecological thinking in epidemiology (Berkman and Kawachi, 2000). The synthesis of these two motivating factors indicated that neighborhoods are an important ecological unit that can act as a stressor, or complex of stressors, to affect health. Suggestions of causal links between neighborhoods and health have long extended beyond the stress model to include
The socio-spatial neighborhood estimation method
With these conceptual and operational issues in mind, we endeavored to develop an approach to define (bound) neighborhood areas within a study of stress and health. This section reports the context of the larger study, the goals of the method, the way the method was conducted, and the data and analysis used to evaluate the method.
Findings
Table 2 shows the ICCs estimated for all measures and neighborhood types. The SNEM neighborhoods produced better ICCs for subjective measures in all cases except for the social embeddedness score, where block groups performed better. Objective measures results also indicate good performance by SNEM neighborhood units relative to the other two neighborhood types.
Table 3 reports the results of the criterion-related validity assessment. The boldface z-scores indicate statistically significant
Discussion and conclusion
The ICC comparison supports our hypothesis that the SNEM approach results in socio-spatial areas that are more reflective of the neighborhood concept. We should note, however, that the geographical basis for the neighborhood perception items—asking respondents to rate an area a few blocks around their home—probably created some error in those aggregated measures. However, the fact that our SNEM ICCs on the two objective measures were much larger than those for the other units of analysis is
Acknowledgments
The authors wish to acknowledge the contributions of Peter Dana and Joseph Forrest on GIS work related to this project, including imagery, maps, geocoding, and analysis. Tasanee Walsh also contributed earlier analysis that led to the work reported here. Steve Owen added statistical suggestions about neighborhoods and sampling as well as about validation. This work was supported by Grant P50 CA10563 from the National Cancer Institute, which funded the UTMB Center for Population Health and Health
References (51)
- et al.
A brief observational measure for urban neighborhoods
Health & Place
(2001) - et al.
Zone design for environment and health studies using pre-aggregated data
Social Science & Medicine
(2005) The need for the ‘‘new health geography” in epidemiologic studies of environment and health
Health and Place
(2007)- et al.
Framing the biosocial pathways underlying assumptions between place and cardiometabolic disease
Health & Place
(2008) Neighborhoods and health: where are we and where do we go from here?
Revue d'Epidémiologie et de Santé Publique
(2007)- et al.
Constructing data zones for Scottish Neighbourhood Statistics
Computers, Environment and Urban Systems
(2007) - et al.
Neighbourhood effects on health: does it matter where you draw the boundaries?
Social Science & Medicine
(2008) - et al.
Neighbourhood variations in child accidents and related child and maternal characteristics: does area definition make a difference?
Health and Place
(2008) - et al.
Modifiable neighborhood units, zone design and residents’ perceptions
Health & Place
(2007) - et al.
Physical activity in children: does how we define neighbourhood matter?
Health & Place
(2010)
From place-based to people-based exposure measures
Social Science & Medicine
Allostatic load in an environmental riskscape: the role of stressors and gender
Health & Place
Establishing the soundness of administrative spatial units for operationalising the active living potential of residential environments: an exemplar for designing optimal zones
International Journal of Health Geographics
Disentangling the relative influence of built and socioeconomic environments on walking: the contribution of areas homogenous along exposures of interest
Social Science & Medicine
Neighbourhood influences on health in Montreal, Canada
Social Science & Medicine
The spatial dimensions of neighborhood effects
Social Science & Medicine
Small area inequalities in health: are we underestimating them?
Social Science & Medicine
Comparing the effects of neighbourhood characteristics on all-cause mortality using two hierarchical areal units in the capital region of Helsinki
Health & Place
A historical framework for social epidemiology
The concept of neighborhood in health and mortality research
Sociological Spectrum
Individual and contextual risks of death among race and ethnic groups in the United States
Journal of Health and Social Behavior
Perspectives on neighborhood and community: a review of the literature
Social Service Review
Qualitative GIS: A Mixed Methods Approach
Concerns about petrochemical health risk before and after a refinery explosion
Risk Analysis
Putting people into place
Demography
Cited by (45)
Developing a granular scale environmental burden index (EBI) for diverse land cover types across the contiguous United States
2022, Science of the Total EnvironmentCitation Excerpt :A contribution of our study is the development of a national environmental burden index (EBI) that is available and processed at the census tract level, a geographic level finer than that of other cumulative environmental burden indices, with tracts functioning as proxies for communities. Health studies have used census tracts as such and found associations with health outcomes at this geographic level (Cutchin et al., 2011; Owusu et al., 2018; K. Wang and Immergluck, 2018). Census tracts are often the geographic unit of choice for federally targeted interventions in impoverished areas (U.S. Library of Congress, 2020).
A multilevel analysis of the drivers of household water consumption in a semi-arid region
2020, Science of the Total EnvironmentThe body language of place: A new method for mapping intergenerational “geographies of embodiment” in place-health research
2019, Social Science and MedicineStress and health behaviors as potential mediators of the relationship between neighborhood quality and allostatic load
2018, Annals of EpidemiologyCitation Excerpt :Neighborhood-level data were generated by matching U.S. Census 2000 block level data with TCSHS neighborhood boundaries (n = 48), defined by the sociospatial neighborhood estimation method, which incorporates (1) street patterns, (2) residential patterns including housing types and densities, (3) nonresidential land use, (4) landforms including barriers to passage and interaction, and (5) geographic spread. Details on data collection and the sociospatial neighborhood estimation method have been described in prior work with this data set [15–17]. Neighborhood quality was operationalized with three measures: neighborhood socioeconomic status, perceived neighborhood quality, and observed neighborhood quality.