We extend the work of these investigators by further exploring the breadth of medical conditions for which OBGs provide care to their elderly patients and by examining the degree to which OBGs in both rural and urban areas care for nongynecologic conditions. Our findings can offer some understanding of the potential impact of legislation to designate OBGs as primary care providers for elderly women.
Methods
Data Sources
We used the 1994 Washington State Medicare Part B claims file as the data source for our study. This database, part of the National Claims History File of the Health Care Financing Administration (HCFA), is an administrative data set that captures diagnostic and therapeutic information about services billed to Medicare. The Medicare Part B file contains a series of line items with each representing a discrete billable service for a Medicare beneficiary. These line items included identifiers for the patient receiving the service, the physician providing the service, the diagnosis coded, and the date of the visit. Items submitted by physicians include a unique physician identification number (UPIN) used to designate the specialty of the physician providing these services.
Physician Sample
All physicians practicing in the state of Washington and submitting Medicare claims for patient visits in 1994 were eligible for the study. We focused on the approximately 80% of OBGs who were participating in Medicare Part B. A variety of subspecialties including 5 medical (dermatology, cardiology, gastroenterology, pulmonology, and rheumatology) and 4 surgical (general surgery, orthopedic surgery, otolaryngology, and urology) were selected for comparison. Family practice and internal medicine were included in the descriptive analysis of visit frequency but not in the domain analysis, because all of primary care for elderly women was considered within the domain of these 2 specialties.
Information from the HCFA UPIN National Directory, the American Board of Medical Specialties (ABMS) database, and the American Medical Association (AMA) masterfile were used to link specialty information to the UPINs in the Part B Medicare file. Physicians were designated as a specific specialty type if both the ABMS certification and the primary self-designated specialty captured in the AMA masterfile were the same. In cases where these differed, the physicians were excluded. Including only those physicians who had the same specialty of training as their reported specialty of practice ensured accurate assignment of specialty.
Practice Location
Physicians were designated as practicing in rural or urban areas based on the ZIP codes of their practice addresses. ZIP codes were assigned as rural or urban on the basis of their proximity to a hospital classified as such by the Washington State Office of Rural Health of the Department of Health. The 5 physicians whose practice addresses were unknown or were in both rural and urban areas were excluded from the practice location analysis.
Patient Visits and Sample
We aggregated all outpatient physician services (eg, diagnostic tests and procedures) provided to an individual on a single date by the same provider into medical encounters. We used the current procedural terminology (CPT) and the HCFA common procedure coding system (HCPCS) for these services to determine whether they involved face-to-face contact with a physician. Those encounters that included such contact were considered visits. Within each visit, we chose one face-to-face line item—the index line—to identify the primary diagnosis for that visit. This index line either contained the evaluation and management code or, in cases without such a code, the face-to-face line item with the highest allowable charge. The International Classification of Diseases, Ninth Revision (ICD-9) diagnosis codes for these index lines were used to identify the primary diagnosis for each visit.15
We included women patients of the study physicians who were aged 65 years and older, enrolled in Medicare Part B, and alive throughout 1994. We excluded the 15% of patients enrolled in a managed care plan during the year to select elderly patients with unrestricted access to physicians in any specialty and because patient visit data were not available for those enrolled in those plans.
Designation of Specialty Domain
To interpret the claims-based diagnoses for each visit, we used diagnosis clusters to collapse the ICD-9 system into 120 groupings of individual diagnostic codes.16,17 The codes in each group share similar pathophysiologic characteristics that often receive similar management. Two physicians reviewed the diagnosis clusters and nonclustered individual ICD-9 diagnoses to identify conditions traditionally cared for by physicians in each specialty. These diagnoses were designated as “in domain”; all others were designated as “out of domain.” Domain assignments were then reviewed and revised by the Medicare Carrier Advisory Committee, which was composed of specialists including all of those represented in our study.