Robert L. Phillips, Jr, MD MSPH
Michael L. Parchman, MD
Thomas J. Miyoshi, MSW
Geographically locating patients to understand access to care and potential influences on health is not a new concept in primary care. However, it is only the more recent advances geographic information systems (GIS) that have made this process more accessible and robust for primary care. In this paper, we describe briefly some key steps in the integration of GIS in primary care research, and summarize an effort to use GIS for improving access to a community health center (CHC). Given the relative universality of such data collection from CHCs nation-wide and recent political commitment to doubling the capacity of CHCs over the next five years, we suggest options for developing centralized processes for evaluating CHC service areas and local, unmet health care access needs.
Dr. Curtis Hames, a family physician, began practice in Evans County, Georgia in 1947. Early in his clinical career, he had a pilot take aerial photographs parts of the county so that he could map the patients in his practice. Dr. Hames did this in conjunction with a systematic collection of family histories as well as biologic and drinking-water samples from his patients. This led to a lifetime of research in Evans County, Georgia on geographic differences in health and the potential etiologies of those differences.1 In 1977, Farley, Boisseau and Froom began using patient and population mapping to do spatial analysis and mapping of community access to care in relationship to their clinic.2 For the next two decades, primary care remained relatively untouched by geographic analysis. Much more recently, GIS has been used to examine issues important to primary care, including access to care.3-6 We will use the findings from one of these studies to illustrate a method for understanding primary care service areas in relationship to other population data that can suggest interventions to improve access to care. We will also suggest an opportunity to apply this method very broadly and rapidly to improve access to primary care nationally.
Access to care is a critical determinant of health. The federal government helps assure access to health care services by supporting a variety of health centers across the United States. The Community Health Center Program is a Federal grant program funded under Section 330 of the Public Health Service Act to provide for primary and preventive health care services in medically-underserved areas throughout the U.S. and its territories.7 In 1998, 670 CHCs provided care to 8.3 million people including 3.3 million without insurance. CHCs are assigned a target service population designated as Medically Underserved Areas or Populations (MUA/P) based on socioeconomic Census data. To improve access to health care, it is crucial to monitor how access varies across geography and subpopulations, which has proven to be difficult.8, 9 In 1999, one of us (Phillips) tested whether GIS could be a tool for looking at access across geography and populations relative to a CHC. GIS was used to analyze the actual service area for a CHC, which we compared to its target service area (based on its MUP designation) and to population data to identify new areas with health care underservice and a need for improved access.1
Defining the service area for this single CHC involved the geocoding of patient records for a full year and weighting these records by numbers of visits (94% accuracy for geocoding). These data were then appended to census-derived geographic shape files. and the Griffith commitment index was used to define the actual service area at the census block group level. ArcView 3.1 (http://www.Esri.com/) was used to generate maps. This process revealed some surprising differences between the actual and target service areas.[figure 1] Nearly half (47 percent) of the census block groups in the actual service area were outside of the target service area. To measure access to care in the same county, we constructed an access variable from other variables in the 1998 Boone County Health and Human Services Needs Assessment survey, a county-wide survey sufficiently sampled to permit analysis at the census-tract level.10 Based on research by Anderson and Aday, we classified individuals as having poor access to health care if (1) they had no health insurance or regular source of health care, used the emergency department as their usual source of care, or had Medicaid/Medicare insurance coverage; and (2) they reported a time during the past 12 months when they needed to see a physician or dentist but for some reason could not (reasons included cost, transportation, physician/dentist would not accept Medicaid/Medicare) or indicated that cost had kept them from filling at least one prescription in the past 12 months.11 These data were likewise geocoded, mapped by census tract, and overlaid with the derived, actual CHC service area which permitted spatial analysis of population access to care relative to the CHC.[figure 2] This analysis suggested that specific census tracts could benefit from outreach efforts or might be candidates for a satellite clinic. A second measure of access for the target population was also developed using the location quotient (LQ) of target households, that is to say, those below 200% of the federal poverty level. Data from the county survey were also used for this evaluation because the 2000 decennial census had not yet been completed. The LQ we used is the proportion of the total clinic patient visits from target households in a census tract divided by the proportion of the total Boone County population in the same household income range residing in that census tract. Data used to calculate the numerator came from 1998 clinic records; results from the 1998 county assessment survey provided the basis for the denominator. This LQ reveals the spatial distribution of impoverished areas with relatively higher and lower use of the CHC.[figure 3]
The analysis of this CHC’s service area and its mapping in relation to measures of poor health care access in the county has proven to be quite valuable. This information has supported an expansion of federal funding for the Center’s functions, led to further research on vulnerable populations that use the clinic, suggested geographic areas that could be targeted for service marketing or new clinic satellites, and offers opportunities for community oriented primary care interventions. This project has led us to begin testing these methods in rural and urban CHC networks. Expanding this effort to networks of clinics in both rural and urban locations will help us refine protocols for mapping service areas for CHCs in diverse geographic settings.
Given that CHCs regularly report specific data to the Uniform Data Set (UDS) as a requirement of federal funding, most Centers have data systems that make replication of our mapping process possible. Actual service area delineation requires only address data linked to some other patient identifier (even a dummy variable to replace actual patient identifiers) to permit weighting of visits. The data collected for the UDS includes a standardized percent-of-poverty which permits numeration of LQ analyses. Census 2000 data will permit analyses of socioeconomic indicators at the census tract level, permitting denomination of LQ calculations, and may permit some proxy measure of access to care, as well. The UDS also captures demographic data and most clinics capture diagnosis data, which would allow clinics to do demographic and/or disease-specific analyses to develop interventions and engage communities to improve health outcomes.
This type of service to CHCs and communities should not only be propagated but should become a centralized function, supported by the same federal agencies that currently support CHCs. Since the UDS collects data from CHCs regularly, once protocols are developed and validated, it would be fairly easy to automate the geocoding and analysis of CHC service areas. A second automatic function could be to relate service areas to local health care access and even LQ maps. With a little development of web-based GIS tools, like ArcIMS, CHCs might even be able to do other spatial analyses of their geocoded patient database relative to other national or local data. With the Bush Administration’s effort to double the capacity of Community Health Centers over the next five years, the relatively minor investment in developing such a tool might make that investment much more efficient.
We would like to thank Dr. Edward Kinman for role in producing the figures for the original, referenced publication.
Dr. Robert Phillips is the Assistant Director of The Robert Graham Center: Policy Studies in Family Practice and Primary Care (part of the American Academy of Family Physicians) and an assistant professor in the Department of Family Medicine at Georgetown University.
The Robert Graham Center
2023 NW Massachusetts Ave
Washington,
DC 20031
202-986-5708
bphillips@aafp.org
Dr. Michael Parchman is an Associate Professor in the Family and
Community Medicine Department at the University of Texas-San Antonio.
parchman@uthscsa.edu
Thomas Miyoshi is a Senior Instructor in the Family Medicine
Department at the University of Colorado.
Thomas.Miyoshi@UCHSC.edu