I. ’Placing’ myself in the field of Geography
I transferred into a Geography PhD program after earning my BA in History (U.C. Berkeley) and a MA in Political Science (UChicago) and completing a year of the clinical audiology (doctorate, Au.D) program. My dissertation project studies how health policies shape health provider availability, with emphasis on audiologists. I came into this project as a clinical audiology (Au.D) student during my clinical observation hours, watching audiologists carefully limit the battery of tests they ran for each patient to what was covered under insurance. The health care policy constraints on audiologists’ practical autonomy and pay came through in class discussions about low reimbursements and reliance on hearing device sales to maintain practice profitability in the absence of Medicare, Medicaid, and private plan coverage of follow-up audiologic care beyond physician-referred assessments.
Since joining the Geography PhD program, I have taken coursework spanning the depth and breadth of GIScience, Spatial Analysis, Health Geography, Biostatistics, and Spatial Epidemiology. Simultaneously, I was reading very widely to get a better mooring and address the questions: what is geography? How does my work fit into health geography? What are the theoretical underpinnings of the methods used in my subfield? How do they differ from health economics or health services research? That reading also helped a great deal with my qualifying exams and drafting my dissertation proposal.
And now that I am in a health policy fellowship with scholars from public health, epidemiology, health policy and management, biochemistry, and so on, I’m more acutely aware of Geography’s status as a synthesis discipline. As a geographer, I am expected to engage social/political theory and methodologies across subfields and disciplines. My introductory coursework aimed at familiarizing students with Geography’s subfields- physical geography, human geography, and geographic information science (GIS). The latter is both set apart from and embedded within physical and human geography. Anyway, from the beginning, the message was clear: to be a geographer is to be engaged beyond your niche interests. Unfortunately, this interdisciplinarity is not often reciprocated, which makes the work of “translating” knowledge across disciplinary boundaries necessary.
For this reason, I took an interest in spatial epidemiology. Methodologically, it is where Geography and Public Health [specifically Epidemiology- a Greek compound word comprised of the stems: epi, (‘upon, among’), demos, (‘people, district’), and logos (‘study’), or simply the study of the distribution of disease in the population] meet. However, spatial epidemiology is a small subset of public health. Most epidemiologic studies do not address the special considerations of spatial data, even as they use spatially-referenced data. Spatial data should be handled with care and consideration, given that they violate core assumptions of “traditional” statistical methods- most especially independence of values. I have also noticed that public health researchers use datasets (such as indices) that have spatial autocorrelation built into them.
For example, kriging and kernel density estimation are wonderful approaches for estimating values for continuous values (e.g. disease prevalence or pollution levels) across a study area based on a handful of sampling locations. Essentially, kriging or kernel density estimation can be used to convert point data to a raster surface with continuous values that can then be assigned to the centroid of an administrative areal unit, like a city ward, a census tract, or a county. The problem that arises is that the generated values are spatially autocorrelated, and any spatial regression model or other form of spatial statistics using that variable needs to account for the non-independence of those values. It goes back to the 1st Law of Geography: “All things are related, but near things are more related than far things” (Tobler, 1970). Just as important is the 2nd Law of Geography: spatial heterogeneity and the importance of context (Goodchild, 2004).
II. Reaching Across Disciplinary Boundaries
Still, when I try to explain these things, I feel as though I am too “in the weeds.” Spatial (or spatiotemporal?) thinking- or thinking about phenomena and places (even places-as-phenomena) as processes requires a different orientation and attentiveness to context and contingency.
Frequently, when we talk about differential levels of exposures to pollutants in the places we live, work, and play, we are talking about the more macro, policy-level processes of allocating site locations for pollutant-generating facilities and the subsequent exposures (air, water, soil) borne by residents close to these sites. Consider a case study of environmental racism: approximately 70% of superfund sites are within a mile of HUD public housing (EPA-HUD, 2017). This disproportionately affects low-income, people of color, and disabled residents, whose exposures to residual industrial runoff exacerbates pre-existing health outcomes driven by poverty and associated psychosocial stress, which undermines immune functioning and increases susceptibility to illness. That’s just one example.
III. Framing Matters
Let’s take the popular catchphrase, “your zip code determines your life expectancy.” It’s catchy, right?
Well, there are several problems with it, however. This is a classic example of what is called the “ecological fallacy”- extrapolating from population-level data to individuals within the population. Where one lives is not a simple matter of destiny, and places are not homogeneous.
Moreover, zip codes (in the US) are particularly subject to a problem that geographers know as the “Modifiable Areal Unit Problem” (MAUP, for short). In 1984, geographer Stan Openshaw wrote: “the areal units (zonal objects) used in many geographical studies are arbitrary, modifiable, and subject to the whims and fancies of whoever is doing, or did, the aggregating.”
We would do well to understand that zip codes reflect US Postal Service delivery/service routes. They may map more closely to road networks than they do to populations. Below is a sampling of the problems associated with zip codes:
- Zip codes are not contiguous to city or county boundaries, which limits the comparability of analyses
- Nor are zip codes are not coterminous with school districts, water districts, or voting districts
- Alternatives: census block groups, wards, planning districts
- Zip codes vary quite a bit in size and population density, making them unsuitable- to a degree- for comparative analyses
- Year-to-year comparisons of zip codes are methodologically unsound because zip codes’ geographical areas/boundaries can and do change
- e.g. road construction, demographic shifts (depopulation, population growth)
[If you’re curious about the problem with zip codes, I wrote about that not long ago!]
Bottom line: There is a complex interplay between where we live and the health outcomes borne by our communities that is mediated by federal, state, and local laws, policies, and regulations and their enforcement. For example, a neighborhood reflects historical and political processes that shape the distribution of say, home loans with favorable terms (e.g. redlining) and their downstream effects, such as home valuation metrics that implicitly consider the racial demographic makeup of a neighborhood. These historical and current policies and practices have downstream effects. Home values today are influenced by historical patterns and practices of racist discrimination in mortgage lending (redlining) and restrictive covenants (formal and informal agreements that enforced racial homogeneity in neighborhoods, to the exclusion of Black, Latinx, and others). This has major implications for tax policies and resources in schools, which in turn affect prospects for residents. In that sense, a “neighborhood” can be understood as both a place and a social process.
The point here is that space should not be treated as null, or a mere container. That is an impoverished understanding of space that cannot apprehend the emplaced phenomena under study. While most quantitative spatial analytic approaches implicitly assume that space is a container, your interpretation of the results—or even your formulation of the variables does not need to stop there. As researchers studying the health of the population and generating the evidence base that potentially informs policy-making, we have a responsibility to take care in defining the problems, research questions, and approaches we employ.
At best, the research questions are relevant and the methods and units of analysis reflect a context- and process-based understanding of the mechanisms driving exposures affecting population health. Unfortunately, the reality is that our research questions may be delimited by the availability and accessibility of the data, as well as their actual structure (e.g. point data vs. aggregated area-based estimates). Given those constraints, it really is about doing the best we can with what is available to us.
Bloomberg, Bureau of National Affairs, (March 2017). “Majority of Superfund Sites Near Low-Income Housing,” https://www.bna.com/majority-superfund-sites-n73014450645/ (accessed 4 Jan 2019)
Goodchild, M. (2004). The Validity and Usefulness of Laws in Geographic Information Science and Geography. Annals of the American Association of Geographers. 94(2). 300-303. DOI: https://doi.org/10.1111/j.1467-8306.2004.09402008.x
Goodchild, M.F. (2004) GIScience, geography, form, and process. Annals of the Association of American Geographers 94(4): 709–714.
Openshaw, S (1984) Ecological Fallacies and the Analysis of Areal Census Data. Environment & Planning A. 16(1) https://doi.org/10.1068%2Fa160017
Openshaw, S., (1983). The modifiable areal unit problem. CATMOG (38). [URL: https://www.scribd.com/document/343456450/Openshaw-1984-MAUP ]
Tobler, W. (1970). A computer movie simulating urban growth in the Detroit region. Economic Geography. 46, 234-240