A statewide stratified, random sample of diabetes-related hospital discharges for Medicare and Medicaid recipients was reviewed to assess principal diagnosis coding validity. After reabstracting, only 60 percent of the sample were found to be unequivocally correct. In another 23 percent, the correct principal diagnosis could not be determined with certainty. Furthermore, ten percent of the cases where diabetes mellitus was coded as another diagnosis were ambiguous as to whether diabetes actually was the principal diagnosis. The potential impacts of these ambiguities on measures of admission rates, average length of stay, and hospital payment are examined.