Harnessing Geographic Information Systems (GIS) to Enable Community-Oriented Primary Care

Methods

This study used informant interviews and focus group techniques to evaluate a collaborative community-based project between Baltimore Medical System, Inc. (BMSI) and the Robert Graham Center for Policy Studies in Family Medicine and Primary Care. It was approved by the Georgetown University Medical Institutional Review Board.

Study Setting, Community Project Description, and Data Sources

We extracted from practice management software a comprehensive dataset of the nearly 30,000 patients who had visited a system of 7 federally qualified CHCs in 2003. At the time the system of clinics employed more than 40 primary care providers in the service of indigent patients in urban and periurban Baltimore, Maryland. We cleaned and geocoded all clinical data representing unique patients during a single year period, achieving an 80% geocoding match rate. Using ArcView GIS software (ESRI, Redlands, CA), we integrated the geocoded clinical data with US Census data and other publicly accessible spatial information to create simple analytic maps and outputs in a variety of formats. To reveal neighborhood utilization patterns for each clinic and to create individual clinic and system-wide service area maps, we reviewed possible spatial techniques, selected a modified Griffith's commitment index and evaluated methods previously described in the limited primary care literature on the subject (see Figure 1).¹¹ The index measures the degree of patient utilization of a clinic from individual geographies—in this case all census block groups or tracts containing patients from the clinics that were studied. Tracts were ranked by index in descending order and were included in the overall clinic or system-wide service area only until 60% of all patient data were included. Maps of the designated MUA for a clinic were also generated (Figure 2) to demonstrate differences between federally intended and utilization-derived (actual) service areas. We additionally plotted patient and subpopulation distribution maps and calculated area penetration rates for each clinic (see Figure 3), which were defined by the number of yearly clinic users per total population in each census tract according to the 2000 Census. Our methods are described in greater detail elsewhere.¹⁰ These maps were then presented to BMSI clinicians and leaders as a series of focus groups and in-depth interviews designed to evaluate the utility of analytic mapping in a primary care setting and the potential for and barriers to broader application.

• Download figure
• Open in new tab

Figure 1.

Baltimore Medical Systems, Inc.'s’ aggregate service area. Using billing and administrative data assigned to clinic, this map shows the service area (census tract) of 4 BMSI clinics and their overlap.

• Download figure
• Open in new tab

Figure 2.

This map demonstrates the federally designated medically underserved area that a single Baltimore Medical Systems, Inc., clinic receives federal funding to serve, with quarter-mile radii from clinic marked out to 2 miles. Discrepancies between maps of these “ideal” service areas and the actual service areas (such as that in Figure 1) were cited by BMSI executives as informative and surprising; they later used the actual service area information to change a relocation decision based previously on medically underserved area alone.

• Download figure
• Open in new tab

Figure 3.

Census tract population penetration rate for a single Baltimore Medical Systems, Inc., clinic. Revealing dense utilization (as high as 30% of the population) by persons in the census tracts most proximal to this clinic, these maps were cited as critical to strategic planning for future expansion and service delivery. GIS, geographic information systems.

Results

With minimum explanation and a brief iterative exchange, BMSI administrators and their information technology support team were able to produce a comprehensive, 1-year, patient-level database for analysis. Geocoding yielded an 80% first-pass match. Service area maps revealed previously unrecognized range and overlap of individual clinical services (Figure 1), whereas MUA/service area match and area penetration densities varied widely by clinic (Figures 2 and 3). As qualitative reporting revealed (below), dynamic mapping and static maps—such the samples seen in the figures—helped BMSI leaders identify and strategize around unrecognized patterns of clinic use, including unexpected use by remote populations. They also revealed the geographic dimensions of BMSI's primary care services and the extent to which the communities they served depended on their clinics for primary care services.

Initial qualitative analysis revealed the following 7 central themes and perceptions:

1. Mapping and spatial analysis of clinical data resulted in engagement of busy administrators and clinicians. Responses were enthusiastic regarding the potential of analytic mapping to help guide CHC strategic planning and resource allocation.

2. Participants reported enhanced community comprehension as a result of analytic mapping and were, in fact, better suited to interpreting their mapped clinic data than the investigators; interviewees stated a ready understanding of health service area and penetration maps. Many offered suggestions about how to make the maps more easily interpretable that neither presenting investigator had considered. Both investigators also noted that participants could easily grasp the limitations of mapped data and were quick to recognize the perils of overinterpretation of map data in isolation. Participants from the Clinical Operations Committee were visibly engaged in the presentation portion of the focus group, interrupting the authors repeatedly with queries and suggestions. Both investigators noted that committee members had visible epiphanies (“a-ha moments”) when they saw certain maps. After presenting aggregate BMSI service area maps (Figure 1), the presenters were interrupted by multiple participants at once, as if the maps had unveiled a mystery. During the 6 months after our interviews, participants requested a number of additional maps to support a BMSI application to expand that clinic.

3. Participatory interpretation of mapped clinic data was essential and was deeply enriched by provider and administrator knowledge of their patient population. Participant knowledge was essential to and revealed through the analytic map interpretation process. As noted above, interviewees applied their experience and local knowledge to interpret service areas. However, multiparty interaction between technical staff, administrators, clinicians, and community members allowed the unique knowledge and expertise of each—whether in data, neighborhood geographies, community history, or other areas—to capture the real power of mapped data. Energy was highest in the group settings, and investigators were struck by how essential local knowledge was to interpreting clinical and community maps relative to tables, graphs, and other forms of data frequently reviewed in these settings. In both focus groups, considerable and far-ranging knowledge of socioeconomic status and social capital among participants was revealed amid animated discussions of why maps showed variable service penetration by African-Americans in specific neighborhoods.

4. Mapping existing and previously dormant practice management data in this fashion galvanized new ideas about data use for strategic planning and population management. CHC leaders and staff quickly and enthusiastically generated a list of additional spatial analyses that they desired to improve their operations. Further, this phenomenon was not limited to selected individuals; staff from each administrative and clinical unit of the CHC network was able to identify a potential analytic mapping project of value to them. The participatory process generated numerous practical applications for analytic mapping that the investigators had not considered.

5. Mapping was identified as having the capacity to support clinic financing and resource allocation. Leaders and clinicians were most enthusiastic about using analytic mapping to support the expansion of grant applications and clinical operations and to impact the effectiveness of patient outreach.

6. Financial and technologic issues were perceived barriers to project sustainability. Many participants cited concerns that the expense and manpower involved in analytic mapping at the desktop would limit their long-term use of this technology without continued outside assistance. Specifically, they lamented their limited information technology infrastructure and staff expertise in the use of analytic mapping software.

7. Achieving economies of scale across clinics and with foundation or federal support and capturing the power of dynamic mapping through web technology were reported as the best ways to overcome the barriers mentioned above. Participants noted that their likelihood of using clinical data mapping would be far greater if overworked staff had immediate access to their data and the power of GIS when they had a specific question related to mapping. Participants also noted that such a tool should be flexible enough to incorporate other community and population data and to expand into other mapping applications useful to the primary care enterprise.

Investigators further sorted these themes into 3 principal domains of findings: (1) perceptions of primary care implementation of GIS (mostly positive), (2) barriers to further implementation, and (3) possible solutions (see Table 1 for examples and Figure 4 for a summary schematic).

• Download figure
• Open in new tab

Figure 4.

Perceptions, barriers, and possible solutions to increase primary care geographic information systems adoption. BMSI, Baltimore Medical System, Inc.; CHC, community health center.

Discussion

The profound effects of neighborhoods and community on health are well documented.¹⁴ Despite this and continuing calls to understand their patients in the context of community and place,^15–18 most clinicians lack the tools to evaluate and display that community context. In this study we tested novel GIS use in an urban primary care network and demonstrated and categorized the value of analytic mapping to primary care outpatient leaders in identifying community context. Readily available primary care clinical data were easily combined with population data within a GIS environment to understand communities, and clinical leaders affirmed the value of this process for population-based care and planning. Outpatient clinic leaders, clinicians, and community board members in our study were engaged and galvanized by the power of mapped data to enhance community care, and they easily identified applications of analytic mapping that would improve their clinical operations.

Any primary care setting or clinical enterprise could potentially benefit from the techniques described here and previously, in particular the expanding primary care safety net.^19,20 Clearly articulated statements about the value of these tools to clinic and community leaders were couched against the need for less technical and time-intensive processes—processes that could permit secure, interactive queries of mapped clinic and population data, because there is a dearth of tools able to inform primary care leaders about access and utilization patterns at the practice and community levels. On a national level, federal funding initiatives to expand health information technology, regional health information organizations, and uniform data set collection in primary care safety nets are encouraging opportunities for diffusion of analytic mapping in the primary care setting.²¹ Tools are needed more than ever to guide the continued expansion of federally qualified health centers and to identify gaps in safety-net coverage A secure, web-based mapping tool that is broadly implemented in outpatient health centers would complement Health Resources and Services Administration efforts to offer all Americans a “medical home” through projects such the Models That Work campaign and the Primary Care Service Area project.^22,23

Whether findings of our exploratory evaluation of a single, large, multiclinic network that provides primary care to an ethnically and racially diverse population represent broader perspectives merits further study, but our early work with an array of primary care sites has yielded similar results. Although we cannot claim to have represented all clinical and administrative leaders from the organization used for our interviews and focus groups, purposeful and strategic sampling was used to ensure breadth. Additional studies in other settings (eg, rural, suburban, single-practitioner practices, multispecialty practices) are needed to confirm the usefulness of these techniques in wide array of clinical settings. Finally, as we learned from the tendency of our informants in this study, it is essential to understand the pitfalls and limitations of mapped data, which risks misinterpretation similar to and often beyond other forms of data display. Expanded use of GIS to inform planning and policy in the clinical setting will require users to consider issues such as data quality, geocoding match rate, the choice of appropriate scale and areal units, as well as issues of causality versus association, and will probably benefit from additional training or partnership with spatial methodologists.

Conclusion

Our longitudinal experience with this CHC network since conducting these interviews includes continued requests for new maps, request for a contractual relationship to continue their production, and specific application of mapping in planning future service delivery and expansion, clinic moves, and community outreach. Leaders at BMSI have expressed the opinion that this exercise would be measurably more valuable if all safety-net services were similarly mapping their communities and could collaborate on regional planning. However, broader application of these methods is limited by the lack of financial and technical capacity in CHCs and other primary care settings, a theme identified by CHC leaders as a barrier to sustainability of our collaborative mapping project. Commercially available technologies now permit geocoding and housing of outpatient clinical data in a Health Insurance Portability and Accountability Act-compliant fashion; user-friendly internet mapping software have improved substantially in recent years. Such software may offer the ability to overcome the barriers to GIS use cited by clinic leaders and to realize GIS economies of scale for a large body of health centers or primary care clinics, expanding the reach of clinical analytic mapping to CHCs nationwide. Healthy People 2010 calls for an “increase in the proportion of all major national, state, and local health data systems… to promote nationwide use of GIS at all levels,” with a goal of 90% use in such systems.²⁴ Policy makers and regional planners supporting CHCs should consider policies and incentives that broaden GIS implementation at the clinical level.