This case illustrates the use of a geographic information system to monitor and analyze spatial patterns of physicians' multiple locations. Physicians practicing a more than one location on a regular basis are considered multiple-location practitioners and amounted to about 16% of instate (North Carolina) physicians during 1992. The pattern of primary locations (Figure 1) and secondary and tertiary locations (Figure 2) coalesce in metropolitan centers such as Asheville, Charlotte, Greensboro, and Raleigh-Durham-Chapel Hill, North Carolina (Albert 1995). However, 15 counties had ratios of primary to secondary and tertiary locations greater than one suggesting that the contribution of multiple locations to nonmetropolitan counties might be substantial (Figure 3). Monitoring the percent of physicians with multiple locations might serve as an indicator of organizational changes occurring within the health care industries during the 1990s. This case highlights data location, acquisition, and assessment, join and relational operators, geocoding and distance calculations, and standard query language. More technical descriptions of these terms can be found within The NCGIA GIS Core Curriculum and Technical Programs. For more information on geomedical applications of GIS see "Synopsis and Bibliographic Resource of Medical-GIS Applications," Albert (1994) or Albert, Gesler, and Wittie (1995).

Data Location, Acquisition, and Assessment

The physician database is from the North Carolina Board of Medical Examiners (NCBME 1992) via the North Carolina Health Professions Data System (Sheps Center for Health Services Research 1992). The North Carolina Health Professions Data System collects and disseminates data on chiropractors, dental hygienists, dentists, doctors of medicine and osteopathy, licensed practical nurses, nurse practitioners, optometrists, pharmacists, physical therapists, physical therapy assistants, physician assistants, podiatrists, practicing psychologists, psychological associates, and registered nurses. From a larger number a fields from the physician database, 22 were chosen for this geographic analysis:

These 18,253 records and 22 fields were in ASCII format (tab-delimited) and 2.3 megabytes in size (Table 1).

These data were from the 1991-92 application of medical license registration with the North Carolina Board of Medical Examiners (1992). Collection is via self report. This raises concerns over potential errors and missing observations within the physician database.

• First, some corrections were made with respect to misspellings and variant spellings (i.e., misspellings Madion and Manor of Marion, North Carolina).

• Second, efforts were made to assess completeness. Eleven of the 22 fields were over 90% complete. The city, county, and state fields for primary, secondary, and tertiary locations were between 98-99% complete. However, the ZIP Code fields for primary, secondary, and tertiary locations were, respectively, 97%, 66%, and 54% complete.

Overall assessment of the database is fair to excellent depending on the combination of fields under review (Table 2). Software/hardware configurations for this illustration include MapInfo® 2.0.3 for Macintosh on a Quadra 840AV with a 16 inch color monitor, Tektronic 220e color printer, and a LaserWriter as peripheries.

Results

This case illustrates join and relational operators, geocoding and distance calculations, and standard query language. The findings are set out according to these headings so as to conform to specific units within Part I of The NCGIA GIS Core Curriculum for Technical Programs.

Join and Relational Operators

Additional attributes such as a physician classification scheme (i.e., the 105 primary specialties in the database were coded into either General and Family Practitioners, Medical Specialists, Surgical Specialists, or Other Specialists), metropolitan status (metropolitan county or nonmetropolitan county), and population size (U.S. Bureau of the Census 1990) were brought into the physician database through join and relational operators; such operators made the following analyses possible. Multiple-location practitioners were less apt to be medical specialists, 27% vs. 32% (p < 0.001), and more apt to be other specialists, 32% vs. 27% (p < 0.001), than single-location practitioners. Further, multiple-site practitioners were moving down the urban system to secondary and tertiary locations indicating a redistribution of physicians to smaller-sized places. For example, 24% of secondary locations and 35% of tertiary locations were in places under 5,000 compared with 15% for primary locations. Fifty-four percent of physicians have urban primary and secondary (including tertiary) locations, 19% urban primary and rural secondary locations, 5% rural primary and urban secondary, and 20% rural primary and rural secondary locations.

Geocoding and Distances Calculations

Observations within the physicians primary, secondary, and tertiary ZIP Code fields were geocoded using a centroid coordinates file (latitude and longitude). Straight-line distances between location orders were found using the program syntax within MapInfo® (Table 3). On average, there were twenty-four miles between primary and secondary locations (N = 1,741), 30 miles between primary and tertiary (N = 292), and 35 miles between secondary and tertiary locations (N = 286). Note that distances between the locations of general practitioners and family practitioners (GP&FP) were less than between other specialists (OS). Further 90% of primary to secondary distances were between 45-50 miles apart; 90% of primary to tertiary and secondary to tertiary distances were over 60 miles.

Standard Query Language

Standard query language (SQL) provides a means to search a database per specific user requests. Here, an accounting of physician hours gained and lost to individual counties is possible because of SQL. For example, a search of the database for "Greene" within "Primary County", "Secondary County", and"Tertiary County" fields retrieves a set of physician records. Remember, the records contain the fields "Primary Hours", "Secondary Hours", and "Tertiary Hours" that provide physicians' hours. The search found that 18.17 hours per week were brought into Greene County, North Carolina, from secondary and tertiary locations (NCBME 1992).

Health Care Policy Implications

The case illustrates the usefulness of geographic information systems to monitor and analyze patterns of health care providers. However, it is important to place these findings in terms of health care policy implications (Albert and Gesler 1997). Some implications include concern over data quality (missing observations and errors), monitoring of health care personnel (consider multiple locations, otherwise under or over estimate personnel resources), access and availability (consider redistribution of physicians via multiple locations), shortage designations (multiple locations change population-to-physician ratios that determine federal shortage designations), and physician recruitment (target multiple-location practitioners for full-time commitment to shortage communities).

References

Albert, D., and W. Gesler. 1997. Multiple locations of medical practice in North Carolina: Findings and health care policy implications. Carolina Health Services and Policy Review 4: 55-75.

Albert, D., W. Gesler, and P. Wittie. 1995. Geographic information systems and health: An educational resource. Journal of Geography 94:350-356.

Albert, D. 1995. Is there a doctor near the house? MapInfo® analyzes health care access in North Carolina. GlobalNews (Summer): 7.

Albert, D. 1994. Geographic information systems (GIS). In: Geographic Methods for Health Services Research: A Focus on the Rural-Urban Continuum, ed. T. Ricketts, L. Savitz, W. Gesler and D. Osborne, 201-206. Lanham, MD: University Press of America.

North Carolina Board of Medical Examiners. 1992. Physician Database, 1992. Raleigh, NC.

Sheps Center of Health Services Research. 1992. North Carolina Health Professions Data System [Brochure]. Chapel Hill, NC.

U.S. Bureau of the Census. 1990. 1990 Census STF 1A, File O, Geography and Tables P1 to P10. Washington, D.C.