Medicare claims databases have several advantages for use in constructing episodes of care for outcomes research. They are population-based, relatively inexpensive to obtain, include large numbers of cases, and can be used for long-term follow-up. However, the sheer size of these claims databases, along with their primarily administrative (as opposed to clinical) nature, requires that researchers take special care in using them. The 10 PORTs using Medicare claims provided information on their approach to several key issues in working with these data, including: 1) identifying the index cases or patient cohorts to be studied; 2) defining the length of the episode; and 3) measuring outcomes. This paper reports the experience and knowledge gained by these PORTs in using these claims to create and analyze episodes of care.