This two-year study addresses the sub-optimal rates of cancer screening in primary care, by examining the relationships between individual physicians and their medical groups in the provision of preventative cancer test screening tests to Medicare beneficiaries. The project breaks new ground by linking individual physician and medical group-level screening rates. Previous interventions to improve cancer-screening rates have focused on patients, their physicians, and broadly defined """"""""organizational"""""""" change. The premise of this proposal is that medical groups, which increasingly dominate the organization of primary care, are an important source of variation in testing rates and provide a key point of intervention. Key research questions are: 1) Variation in Rates: How much do rates of cancer screening differ among group practices and how consistent is behavior across different screening tests? 2) Contribution of Medical Groups: To what degree are apparent differences among physicians in the cancer screening rates of their patients due to systematic differences among their practices? 3) Group Characteristics: What characteristics of medical groups are related to differences in the performance of their member physicians? The study will use a sample of 229,000 fee-for-service Medicare beneficiaries in three Ohio MSAs. The Ohio sites represent common market structures and include a diversity of medical groups. Medicare Part B and Outpatient claims files for the year 2002 will be analyzed to calculate rates for five screening tests (mammography, sigmoidoscopy, fecal occult blood, PAP smears, and PSA tests). We take advantage of newly available linkages that allow us to use existing American Medical Association data to tie physician profiles to their medical groups using Medicare fee-for-service beneficiaries as the study population. Medicare patients will be assigned by plurality visit to primary care providers. The universe of medical groups will be created by pooling the AMA's physician and group files. Providers will be matched to groups by professional address. Multilevel regressions will estimate the separate and joint effects of physicians and their groups. Since the proposed methodology is new, a small grant is proposed to develop and validate the approach and to generate initial substantive results.
