Tools for Monitoring the Health Care Safety Net

Mapping Tools for Monitoring the Safety Net

By Robert L. Phillips, Jr., M.D., M.S.P.H.; Andrew Bazemore, M.D.; and Thomas J. Miyoshi, M.S.W.


Contents

Background
Mapping Safety Net Provider Services and Communities
Data and Tools for Mapping and Monitoring
The Power of Mapping Tools
References

Background

Geography plays a critical role in the relationship between health care safety net providers and the populations they serve. Beyond the quantity and types of services and providers available, their location relative to where their users live is a crucial dimension of access to care. Given the critical importance of geographic factors in this relationship and the power of visual display, data mapping is well-suited to understanding and monitoring the status of the health care safety net. Geographic Information Systems (GIS) permit the combination of complex population, health care services, and clinical data into maps that illustrate these relationships at fine resolution and can help inform decisionmaking regarding care for safety net populations.

Analytic mapping tools are extensively used to map population data and are increasingly being used in a variety of ways to examine the relationship between health services and populations (Farley, Boisseau, and Froom, 1977; Rushton, 1999; U.S. Census Bureau American FactFinder; Melnick, 2002). Although recent efforts have used mapping tools to create primary care rational service areas (PCSAs) nationally (U.S. Dept. of Health and Human Services), to date most of the efforts to use maps to relate health care services to populations have focused on small geographic areas. Such a focus is well-suited to efforts to monitor the safety net.

The utility of mapping tools for monitoring the safety net and the populations it serves depends not only on the data available and the geographic unit of analysis, but also on the extent of leadership, community buy-in, and the ability to garner support and authority from governmental entities (Ricketts, Savitz, Gesler, et al., 1997). Ideally, the process involves both safety net leaders and users in the analytic process, and both qualitative and quantitative assessment of the significance of the findings.

As an example, clinical data from community health centers has been used in conjunction with population data to provide small-area maps that can help in decisionmaking about expansion of services, planning interventions, and defining safety net populations that need services (Phillips, Kinman, Schnitzer, et al., 2000; Phillips, Parchman, Miyoshi, 2001). Similar research efforts are under way for redesignation of Health Profession Shortage Areas as well as in access to care in rural areas (Juarez, Robinson, Matthews-Juarez, 2002; Gesler). With community leadership and governmental support, these tools could be more broadly applied for evaluating and monitoring the health care safety net.

This chapter offers examples of work with community health centers and describes the basic elements needed from all safety net providers to create more comprehensive service maps. It includes maps of population data that demonstrate potential safety net needs, discusses sources of these data, and illustrates how to combine these data with safety net maps to examine the status of the safety net. This chapter suggests how these analyses can be used to focus clinical and policy options, community action, and political will. Lastly, it addresses how mapping tools combined with community leadership and governmental support/authority could become part of a longitudinal monitoring effort.

Return to Contents

Mapping Safety Net Provider Services and Communities

The current Administration's initiative to expand the capacity of community health centers provides an opportunity to test mapping tools as a way to support expansion efforts. For example, Baltimore Medical System, Inc. (BMS), the largest community health center (CHC) network in Baltimore, MD, operates a network of stand-alone and school-based clinics. After receiving approval from the Georgetown University Medical Institutional Review Board and taking steps to ensure that individual patients could not be identified via our mapping efforts, we used BMS' patient records combined with population data from the 2000 U.S. Census to define BMS' service area and to look at its penetration in specific neighborhoods relative to potential need for services.

Partially funded under Section 330 of the Public Health Service Act, CHCs such as those operated by BMS are required to report patient care information regularly and therefore often have fairly complete patient information linked with demographic, disease, and income information that facilitates analytic mapping. Other safety net providers often have at least patient address information in an electronic format. Including data from multiple safety net providers in a given area does not pose additional challenges beyond analyzing data from a single provider.

Health Care Service Areas

BMS' leadership felt that using data on patients to define its service area would be more valid than using data on visits, since frequent users might skew the mapping efforts. Ideally, a service area should reflect the spatial distribution of people using each clinic and the clinic network relative to how those services are organized. We defined the service area as the census block groups with the most users that cumulatively comprised 60 percent of all users (Figure 1) (Ricketts, 2002). Census block groups are the next-to-smallest geographic units defined by the U.S. Census, ideally containing 400 housing units (U.S. Census Bureau, 1994). Sixty percent is a traditional service area definition, but we continue to explore better methods for defining service areas (Griffith, 1972; Cromley and McLafferty, 2002; Ricketts, Savitz, Gesler, et al., 1997).

Using ArcView® 8.1 software, the service areas for each of the four clinic sites that had been in operation for the 1-year period from June 2000 through June 2001 were mapped. Overlapping health service areas were discovered between Matilda Koval Health Center (MKHC) and its sister clinics, Belair Road Family Health Center and Highlandtown CHC, in areas that were expected to fall solely within the MKHC service area. Leaders of BMS were surprised by this finding, and noted that it helped explain why Matilda Koval was perpetually operating beneath its estimated patient capacity while its two sister clinics were very busy (Figure 1).

Conducting similar mapping exercises for multiple safety net providers in a given area would help depict patterns of use by underserved populations and may be even more helpful for identifying geographic gaps in care. For example, if an emergency room's service area is geographically outside of any primary care safety net providers' service areas, options could include creating a new health center or other primary care facility, or outreach efforts to inform the community about other care options in their area. As with BMS, service area mapping may also identify areas of service overlap that can stimulate further investigation about why communities are using (or not using) certain safety net resources.

Service area mapping for health centers also offers an opportunity to compare actual clinic utilization patterns to those presumed under the CHC's designated Medically Underserved Areas/Medically Underserved Populations (MUAs/MUPs). (For information on MUAs/MUPs, go to the chapter entitled Rural Health Care Safety Nets, by David Hartley and John Gale). The comparison of a CHCs actual health service area and its proposed area of service—as suggested by MUAs/MUPs—may offer a way to validate the extent to which MUAs/MUPs effectively reflect underserved communities. Furthermore, service area mapping may help safety net providers to make a case for new MUA/MUP designations or for providing services to areas and populations outside of their designated MUAs (Figure 2) (Ricketts, 2002). Mapping the service areas for multiple safety net providers may also create rational areas for other purposes such as mass vaccination planning, strategic placement of emergency resources, and public health or even bioterrorism planning.

Penetration Rates and Other Measures of Clinic-to-population Relationships

One measure of "market share" or local community use for clinics is a penetration rate. This is obtained by dividing the number of patients coming to the clinic from a given area by the total population of that area. This was done for each of the four clinics at the census block group level. BMS was pleasantly surprised to find that one of their more established clinics was visited by up to one-quarter of the local population. These maps graphically demonstrated to BMS leaders their position as the major safety net provider for this area, despite their close proximity to Johns Hopkins University and its emergency and outpatient services. Given the reliance of the local communities on this clinic, as revealed by the penetration map, BMS leaders felt reaffirmed in their decision to investigate expansion funding for that clinic (Figure 3).

Penetration rates can also be tailored for more focused analysis of target populations. For example, using U.S. Census demographic data at the block level and socioeconomic data at the block group level, it is possible to examine penetration rates for African-American children under age 12 at the block level, and penetration rates for people under 200 percent (of the Federal poverty) at the block group level.

One practical application of detailed penetration rates maps and similar techniques relates to a grant-funded BMS outreach program to increase screening mammography for African-American women age 40 and older. To help the outreach coordinator prioritize the locations of future efforts, we mapped the location of BMS's patients for a single clinic combined with data on the area's population of eligible women (Figure 4). These maps provided the six-person BMS outreach team with a visual relationship between their clinic's "community" of African-American women over age 40 and how it compares to the overall population.

After a short training session using dynamic versions of the map (Figure 4), which allowed team members to initiate their own queries of the data, BMS's outreach team modified its plan. A dependence on mass mailings and presentations at numerous community meetings and churches has been replaced by geographically directed, door-to-door outreach in census blocks with the highest density of the target population:

BMS also has used analytic maps to help inform a potential clinic relocation decision. To assist them, patient locations for several target populations were mapped—including African Americans, children under age 12, and adults over age 65—to understand how a move 2 miles northeast from the present site might affect these populations. Based on what the clinic perceived to be significant clustering of these patients to the south of the present clinic site, they ultimately decided to chose a new building next door to their present site (Figure 5).

These three efforts show how mapping tools can relate clinical data to population data to display penetration of services in small geographic areas, utilization by target populations relative to their population density, and to assess distance-to-care as a factor for making decisions about clinical resources. Future analyses will add bus routes to the analyses to help assess the role of transportation routes as well. These mapping tools could be useful in monitoring or assessing safety net utilization and can help provide insight into how the placement or movement of services might change utilization patterns.

Other Uses for Mapping Tools

Maps of emergency department utilization patterns by surrounding populations can reveal gaps in the primary care safety net and potential neighborhoods with clinical service or resource needs. Analytic mapping can also allow planners to understand patterns of emergency department utilization for primary care services by patients living within easy reach of other safety net facilities. Existing studies on the subject lack analyses of geographic variables that impact utilization and would be augmented by GIS analyses (McCarthy, Hirshon, Ruggles, et al., 2002). With regular data input, analytic mapping could also allow dynamic real time disease surveillance within the safety net. For example, a rise in emergency department use by children with flu-like symptoms could be examined for geographic patterns. Analytic mapping has been used extensively in this manner to examine patterns of elevated childhood lead levels (Roberts, Hulsey, Curtis, et al., 2003), and the BMS clinics already serve as sentinel health data collectors for Baltimore's bioterrorism preparedness program. With a comprehensive mapping program, disease patterns could also be examined relative to primary care safety net service areas. Looking at outbreak patterns relative to clinical resources could offer opportunities for testing, educating, or even treating other patients in target clinics.

The overall picture of the safety net depicted in maps of health service areas, penetration rates, and other aspects of any safety net entity or system are only partially informative in the hands of the analytic mapmaker or policymaker alone. As seen during focus groups at BMS, the power of analytic mapping lies in its ability to galvanize safety net leaders, clinicians, and community board members. Mapping tools enable those interested in the safety net to partner in the process of monitoring and improving it through visualizing disease or health care utilization patterns, seeking explanations for gaps in service provision or discovering inequities among targeted demographic groups.

Return to Contents

Data and Tools for Mapping and Monitoring

Secondary Data

U.S. Census Bureau data are some of the most easily mapped, and many variables are available at the block or block group level. These are available by State, free of charge, at http://www.census.gov/main/www/cen2000.html. This Web site also has links to necessary shape files, the files used to define and create key map elements like borders and roads, for each county and State as low as the block level. These shape files can also be downloaded and used in many mapping software packages. For more information, go to http://www.census.gov/geo/www/tiger/index.html, under the heading Topologically Integrated Geographic Encoding and Referencing system (TIGER). Many maps presented here used shape and population data downloaded from the U.S. Census Bureau. Two of the major limitations of Census data are the lack of information about health and the changes that can occur in the 10 years between censuses.

The Centers for Medicare & Medicaid Services (CMS) are another source of secondary data. These data have both patient and health care provider data and, therefore, are not only useful for mapping service areas, but also can help identify providers who treat low-income populations but are not typically considered part of the safety net. These data can also reveal how patients use multiple types of safety net providers, something that typically cannot be done using individual clinic data. CMS and individual State health agencies have these data at varying levels of specificity; however, public use data files frequently will not allow the disaggregation of data below the State or county level. For the analyses and maps to be of value for monitoring local safety nets, data are needed at much smaller geographic units. CMS and States can permit smaller geographic aggregation of data under special circumstances, although the arrangements to use such data may be complex.

Another source for these data is insurers. In areas with multiple Medicaid insurance providers, this may mean negotiating for data with several organizations. In some States or regions where managed Medicaid is more dominant, these companies may also have data on Medicaid-eligible children or families, even if they are not currently enrolled. In our experience, these companies are sometimes willing to share data because the analyses and maps are instructive for them as well.

These public data have two additional limitations:

  1. Medicare and Medicaid data largely contain information on people over age 65 and children, and they do not represent the uninsured.
  2. The Health Insurance Portability and Accountability Act of 1996 (HIPAA) has created specific protections for identifiable data. HIPAA does not preclude the use of these data and the regulations provide ways to request it and use it appropriately.

Collecting Local Data

Community health surveys can be structured to inform safety net monitoring through the use of mapping tools. For example, Boone County, MO, conducted a sample survey of adults modeled after the Behavioral Risk Factors Surveillance System (Centers for Disease Control and Prevention, http://www.cdc.gov/brfss/) (Boone County Health and Human Services Needs Assessment—1988, 1999). The survey was large enough to permit examination of access to health care by census tract relative to a CHC service area (Figure 6) (Phillips, Kinman, Schnitzer, et al., 2000). Our analytic maps revealed uninsured populations and those with significant barriers to care that did not use the CHC and were well outside of its service area. The leaders of this health center felt that these maps were key to their success in receiving new money for health center expansion. This county, among the top 10 in the country for physician-to-population ratio, encountered difficulty making a case for expansion without the maps.

Many communities periodically study issues relevant to the safety net such as the health and health care access of the population in their community. Particularly if these studies use validated measures, as Boone County did, adding a geographic component is a logical next step and adds value to the results because it locates safety net-sensitive populations and can help target resources. Healthy People 2010 has as a goal that local health data systems be capable of geocoding to allow mapping their data (Office of Disease Prevention and Health Promotion, 2001). By developing the survey sampling process to allow comparisons at specific geographic levels, communities can use this information to understand and depict disparities, reveal gaps in the safety net, and even compete successfully for new resources.

Local Data Collaboratives

Beyond doing their own local surveys and using secondary data, communities increasingly have access to other locally collected data that is relevant to the safety net. For example, the Urban Institute has been supporting an effort to gather local data on social indicators and geographic conditions that affect health and health care. This effort, called the National Neighborhood Indicators Partnership (NNIP), seeks to use data from various local sources to identify potential problems and their solutions and uses an online mapping tool to display these (Kingsley, 1999; National Neighborhood Indicators Partnership). A few States have also begun creating data warehouses and using them to evaluate community health. Data collaboratives or warehouses are an important resource for local efforts to monitor the health care safety net, but their development and maintenance depend on leadership and often on governmental support and authority.

Analytic Mapping Tools

The Centers for Disease Control and Prevention (CDC), the Environmental Protection Agency (EPA), other Federal agencies, and increasingly State and county governments have Web-based or desktop mapping services (National Center for Injury Prevention and Control Injury Maps; National Center for Health Statistics Atlas of United States Mortality; EPA Environmental Atlas; Fairfax County, VA GIS & Mapping—Home Web sites). These are often dedicated to understanding environmental hazards, law enforcement, travel planning, and community services. The infrastructure for analytic mapping may already exist in many communities and could potentially be facilitated for others by State and Federal agencies. Communities that want to do their own analytic mapping have several options for proprietary software, including ArcView®, Maptitude®, MapInfo™, and others. There are even vendors who will custom design and maintain these systems and the data. There are free options, such as EpiInfo™, which is available from the CDC at http://www.cdc.gov/epiinfo/, and can be used by anyone with training.

Return to Contents

The Power of Mapping Tools

We have discussed the use of mapping tools for identifying ongoing community needs and relating these to safety net services, how this reveals opportunities for action, and how one community in Missouri used such information to leverage new resources. Beyond identifying underserved populations and the potential for locating new services, we also discussed how these maps identify rational service areas within the safety net for emergency planning or targeted interventions. Using mapping to help accomplish these tasks and then to evaluate and monitor progress over time can be an additional, powerful tool for the health care safety net. We have also stressed the need for leadership and governmental support and authority to develop and use mapping tools effectively for these purposes.

Mapping tools can also be used for garnering community and governmental leadership, for engendering political accountability, and for leveraging resources and authority. Since maps help people see patterns of disparity and underservice and to locate those patterns geographically within the community, they can be a powerful point of departure for community activism. Health center community advisory board members and staff have responded viscerally to our map presentations and have immediate ideas for how maps could help them. Expanding these tools to an entire local safety net would be similarly powerful, not only serving as tools for revealing, explaining, planning, and evaluating, but also as ways to marshal community leadership. Maps frame problems and successes in geopolitical units for which various political, business, and civic leaders are accountable. We explored this potential in Boone County, MO, by directly applying specific political boundaries to the problem of poor access to health care (Figure 7). The CHC there felt that relating these boundaries to community health care access data would provide additional leverage in obtaining support from politicians. Analytic mapping tools can identify accountable leaders by revealing safety net-related problems contained within specific political boundaries and can give them credit for successes if their response yields demonstrable changes on subsequent maps.

Mapping tools can add real power to efforts to monitor the safety net by providing improved analyses, revealing geographic and clinic/population relationships, depicting rational areas for planning and intervention, and marshaling community activism and political accountability. The tools are already used in many communities for other purposes and could be expanded to include health care safety net evaluations. Working across the many entities that comprise the safety net to collect and manage data collaboratively, conduct appropriate analyses using mapping tools, and use analyses these for planning or interventions will require leadership, support, and authority.

Return to Contents

References

Boone County Health and Human Services Needs Assessment, 1998. Preliminary Findings: Summary Report. Columbia (MO): Boone County; 1999. Available at: http://www.oseda.missouri.edu/special/booneco/rpt9899/. Accessed October 30, 2003.

The CDC National Center for Injury Prevention and Control Injury Maps. Available at: http://www.cdc.gov/ncipc/maps/default.htm. Accessed October 30, 2003.

Cromley EK, McLafferty SL. GIS and Public Health. New York: Guilford Press; 2002. Chapter 9, Analyzing access to health services.

Fairfax County, Virginia GIS & Mapping—Home. Available at: http://www.co.fairfax.va.us/maps/map.htm. Accessed October 30, 2003.

Farley ES Jr, Boisseau V, Froom J. An integrated medical record and data system for primary care, Part 5: Implications of filing family folders by area of residence. J Fam Pract 1977;5:427-32.

Gesler WM, PI (University of North Carolina, Chapel Hill). Geographic accessibility of health care in rural areas. Grant No. R01 HS09624, FY 1999.

Griffith JR. Quantitative techniques for hospital planning and control. Lexington (MA): Lexington Books; 1972.

Hevner AR. Applying data warehousing to community health assessment. Paper presented at: 9th Workshop on Information Technologies and Systems (WITS'99); 1999 Dec 11-12; Charlotte, NC.

Juarez PD, Robinson PL, Matthews-Juarez P. 100% access, zero health disparities, and GIS: an improved methodology for designating health professions shortage areas. J Health Soc Policy 2002;16(1-2):155-67.

Kingsley GT. Building and operating neighborhood indicator systems: a handbook. Washington (DC): The Urban Institute; 1999.

McCarthy ML, Hirshon JM, Ruggles RL, et al. Referral of medically uninsured emergency department patients to primary care. Acad Emerg Med 2002 Jun;9(6):639-42.

Melnick AL. Introduction to Geographic Information Systems for Public Health. Boston (MA): Jones & Bartlett; 2002.

National Center for Health Statistics Atlas of United States Mortality. Available at: http://www.cdc.gov/nchs/about/otheract/gis/gis_atnchs.htm. Accessed October 30, 2003.

National Neighborhood Indicators Partnership—NNIP. Available at: http://www.urban.org/nnip. Accessed October 30, 2003.

Office of Disease Prevention and Health Promotion: Healthy People 2010. Washington (DC): U.S. Department of Health and Human Services; 2001. Available at: http://www.healthypeople.gov/document/HTML/Volume2/23PHI.htm. Accessed October 30, 2003.

Phillips RL, Parchman ML, Miyoshi TJ. Using geographic information systems to understand health care access. ESRI 1st International Conference on GIS in Health Care. ESRI#379-389; Oct 2001.

Phillips RL Jr, Kinman EL, Schnitzer PG, et al. Using geographic information systems to understand health care access. Arch Fam Med 2000 Nov-Dec;9(10):971-8.

Ricketts T. Geography, geographic analysis and the special case of shortage designation in rural health. Paper presented at: A Workshop on Geographic Methods and Measurements and Health Services Research for Vulnerable Populations; 2002; Rockville, MD.

Ricketts TC, Savitz LA, Gesler WM, et al. Using geographic methods to understand health issues. Rockville (MD): Agency for Health Care Policy and Research; 1997 March. AHCPR Publication. No. 97-N013.

Roberts JR, Hulsey TC, Curtis GB, et al. Using geographic information systems to assess risk for elevated blood lead levels in children. Public Health Rep 2003 May-Jun;118(3):221-9.

Rushton G. Methods to evaluate geographic access to health services. J Public Health Manag Pract 1999 Mar;5(2):93-100.

U.S. Census Bureau American FactFinder. Available at: http://factfinder.census.gov/servlet/BasicFactsServlet. Accessed October 30, 2003.

U.S. Census Bureau. Geographic Areas Reference Manual. Washington (DC): U.S. Bureau of the Census; 1994. Available at http://www.census.gov/geo/www/garm.html.

U.S. Department of Health and Human Services. Health Resources and Services Administration. Primary Care Service Area. Available at: http://pcsa.hrsa.gov. Accessed January 13, 2004.

U.S. Environmental Protection Agency Environmental Atlas. Available at: http://www.epa.gov/ceisweb1/ceishome/atlas/. Accessed October 30, 2003.

Return to Contents

Current as of November 2003


Internet Citation:

Phillips RL, Bazemore A, Miyoshi TJ. Mapping Tools for Monitoring the Safety Net. Tools for Monitoring the Health Care Safety Net. November 2003. Agency for Healthcare Research and Quality, Rockville, MD. http://www.ahrq.gov/data/safetynet/phillips.htm


Return to Tools for Monitoring the Safety Net
Data & Surveys
AHRQ Home Page
Department of Health and Human Services