In summary, this study provides evidence that with as few as five diagnosis it is possible to identify some physician group practices that consistently treat patients with more or fewer visits, on average. However, even among group practices that vary considerably in size, location, and organizational structure, there is little deviation from the norm in terms of office visits. A study of more than 30 patients per site using data from both office records and insurance claims is needed, however, to examine the entire spectrum of treatment, including lab tests, special procedures, medications, and hospitalization. Such future studies may exploit the possibilities and avoid the pitfalls describes here to better characterize physicians' practice patterns.