An assessment of the usefulness of the Duke criteria for diagnosing active infective endocarditis.

We evaluated the usefulness of the Duke criteria for diagnosing cases of active infective endocarditis (IE). Patients were identified prospectively over a 3-year period at 54 hospitals in the Philadelphia metropolitan area. Three of us independently reviewed abstracted hospital records and classified 410 patients as definite, probable, or possible cases of IE or as probable noncases. We then applied the Duke criteria to this sample to assess the degree of agreement between our diagnoses and the diagnoses based on these new criteria. Agreement was good to excellent, ranging from 72% to 90%, depending on the case definition used. The sensitivity of the Duke criteria was also good to excellent, varying from 71% to 99%, again depending on case definition used. Specificity was lower (0-89%). We conclude that use of the Duke criteria will result in little underdiagnosis of IE but that it may result in overdiagnosis of IE; therefore, these criteria should be applied prospectively to determine their clinical usefulness.