The Health Care Financing Administration is developing a new method of Medicare patient reimbursement limit setting, based on a quantification of hospital case mix through the AUTOGRP Diagnosis-Related Groups established at Yale University. The reliability of this method is dependent on the diagnosis and procedure data used as input, which should reflect the "principal" condition of the patient. HCFA's source of data is Medicare billing for 1978, some of which contains concurrent rather than retrospective diagnosis and procedure information. Billing data from large teaching hospital are examined with respect to Medical Records data. The data are evaluated, based on the diagnosis and procedure codes and on the groupings (DRGs) presently being used by HCFA; concurrent and retrospective data are found to be widely divergent on both measures. An apparent difference in complexity or extent of resource use is noted, suggesting that the data being used in HCFA's development effort may not fully represent the level of complexity of cases being treated and that reimbursement based on this data may be incorrect.