Increasing attention is being paid to data on geographic differences in population-based rates of use of medical and surgical procedures. To understand these differences and to determine what level of use is appropriate, a method is needed to judge the clinical appropriateness of services. We recently developed and tested such a method in two large, urban geographic areas. Eighty-one medical records from a randomly selected sample of 30 billing entities (46 physicians) who performed upper gastrointestinal endoscopy (UGIE) on Medicare patients were abstracted. Ninety-four percent of physicians who were asked agreed to participate. Reliability testing showed 99% agreement on items abstracted from a subset of records independently reviewed by two abstractors. Based on the abstractions, each patient could be assigned at least one (mean 2.2) specific clinical indication for which UGIE was performed. Using ratings derived from a previously held panel meeting, it was possible to evaluate the appropriateness of the indications for each UGIE.