Negative predictive value of the Duke criteria for infective endocarditis.

With use of new Duke criteria, 405 episodes of suspected endocarditis were previously classified as "definite," "possible," or "rejected" endocarditis. To determine the negative predictive value of the Duke clinical criteria for the classification of suspected endocarditis, chart review and follow-up were performed for the 52 episodes in which the diagnosis of endocarditis was rejected. Three of 52 episodes were reclassified to possible endocarditis; 49 episodes in 48 patients met the criteria for rejected endocarditis. Of these 49 episodes, 31 (63%) had a firm alternate diagnosis other than endocarditis, 17 (35%) had resolution of the clinical syndrome leading to the suspicion of endocarditis with < or = 4 days of antibiotics, and 1 patient had no evidence of endocarditis at surgery. Echocardiograms recorded in 3 patients with rejected endocarditis had evidence of oscillating valvular masses, and blood cultures were positive in 13 episodes; none of these patients had evidence of endocarditis at follow-up. Follow-up or outcome information was available in all 49 episodes. Excluding the 5 in-hospital deaths, mean duration (+/- SD) of follow-up was 39.9 +/- 28.8 months (range 0.5 to 108.0); in living patients, mean time to final follow-up was 56.2 +/- 25.2 months (range 25.0 to 108.0). One patient had possible infective endocarditis at autopsy. No patient in our series whose diagnosis of endocarditis had been rejected had proven endocarditis. Therefore, the negative predictive value of the Duke clinical criteria for endocarditis is at least 92%.