Linking physician characteristics and medicare claims data: issues in data availability, quality, and measurement.

BACKGROUND: Increasingly, investigators are using administrative databases to answer research questions requiring physician characteristics information. This article provides a roadmap for investigators who use Medicare data to answer such questions, focusing on use of the Surveillance, Epidemiology, and End Results (SEER)-Medicare files. METHODS: Three data sources that can be linked to identify physician characteristics-Medicare claims, the Unique Physician Identification Number (UPIN) Registry, and the American Medical Association (AMA) Masterfile-were examined for data availability, linkage rates, and quality. These databases also were used to explore measurement issues regarding physician specialty and practice volume. RESULTS: Over 98 percent of UPINs identified from the Medicare claims could be linked with both the AMA Masterfile and the UPIN Registry. The AMA Masterfile is the best source of sociodemographic and medical training information; the Medicare claims are the best source of practice ZIP code; and the UPIN Registry is the best source of practice organization data. The operationalization of variables such as physician specialty and practice volume is dependent on the specific research question that is being addressed. CONCLUSION: Administrative databases, such as SEER-Medicare data linked to AMA Masterfile or UPIN Registry data, are an important resource for investigators interested in assessing the relationship between physicians' personal and practice characteristics and the content or outcomes of clinical care.

Identifying and measuring hospital characteristics using the SEER-Medicare data and other claims-based sources.

The context in which health care is delivered may have an impact on the outcomes and processes of care. Researchers analyzing claims data that reflect care rendered at many institutions may wish to identify hospital characteristics such as bed size, teaching status, degree of specialization, and case volume to evaluate their influence on outcomes. The authors describe the resources available from administrative data for characterizing hospitals. The article focuses on analyses using the SEER-Medicare-linked data, but has direct relevance for investigators working with other sources of claims data. First, the authors focus on assessment of hospital case volume, which is the principal hospital attribute that can be obtained directly from the SEER-Medicare data files. Second, they assess the information regarding hospital characteristics that can be gleaned from the American Hospital Association (AHA) survey and the annual cost reports submitted to the Center for Medicare and Medicaid Services (CMS) by all certified hospitals. From a comparison of the responses of 4,434 hospitals to 1994 AHA and CMS surveys, the authors conclude that the two data sources contain very similar information, but that for most investigators, use of data from CMS cost reports will be preferable to use of AHA data because of higher survey-completion rates and public availability.