SUMMARY – CORE C There is extensive unexplained geographic and organizational variation in health care utilization and outcomes across the United States. Many health policy experts view the reduction of unwarranted variation to be essential in improving health care quality and reducing costs. Previously, individual heterogeneity of clinicians, hospitals, and health systems has been studied as a possible determinant of such variation. However, substantial variation is left unexplained. We hypothesize that the best way to make further inroads is in the context of clinician networks and that the array of professional relationships between clinicians across the U.S. is the mechanism for the diffusion and dissemination of best-practice (high value, low cost) and worst-practice (ineffective and expensive) care. Unfortunately, it has proven challenging to measure professional networks among U.S. providers. The Methods Core will meet this challenge by (i) measuring relationships from patient- clinician encounters in Medicare claims data using methods that make greater use of the information contained in them than previously, (ii) computing network measures and statistical models involving them for the entire U.S. social network, and (iii) using advanced computing to implement computations on the entire network in feasible time. Advances in statistical methods for modeling hierarchically structured data and causally- defendable comparative effectiveness research will also be made. These methodological innovations support and enhance the five projects, providing them the basis to delve deeper into their various lines of enquiry than would be possible otherwise.