Can structured clinical assessment using modified Duke's criteria improve appropriate use of echocardiography in patients with suspected infective endocarditis?

BACKGROUND

Although echocardiography has been incorporated into the diagnostic algorithm of patients with suspected infective endocarditis, systematic usage in clinical practice remains ill defined.

OBJECTIVE

To test whether the rigid application of a predefined standardized clinical assessment using the Duke criteria by the research team would provide improved diagnostic accuracy of endocarditis when compared with usual clinical care provided by the attending team.

METHODS

Between April 1, 2000 and March 31, 2001, 101 consecutive inpatients with suspected endocarditis were examined prospectively and independently by both teams. The clinical likelihood of endocarditis was graded as low, moderate or high. All patients underwent transthoracic echocardiography and appropriate transesophageal echocardiography if deemed necessary. All diagnostic and therapeutic outcomes were evaluated prospectively.

RESULTS

Of 101 consecutive inpatients (age 50+/-16 years; 62 males) enrolled, 22% subsequently were found to have endocarditis. The pre-echocardiographic likelihood categories as graded by the clinical and research teams were low in nine and 37 patients, respectively, moderate in 83 and 40 patients, respectively, and high in nine and 24 patients, respectively, with only a marginal agreement in classification (kappa=0.33). Of the 37 patients in the low likelihood group and 40 in the intermediate group, no endocarditis was diagnosed. In 22 of 24 patients classified in the high likelihood group, there was echocardiographic evidence of vegetations suggestive of endocarditis. Discriminating factors that increased the likelihood of endocarditis were a prior history of valvular disease, the presence of an indwelling catheter, positive blood cultures, and the presence of a new murmur and a vascular event. General internists, rheumatologists and intensive care physicians were more likely to order echocardiography in patients with low clinical probability of endocarditis, of which pneumonia was the most common alternative diagnosis.

CONCLUSION

Although prediction of clinical likelihood varies between observers, endocarditis is generally found only in those individuals with a moderate to high pre-echocardiographic clinical likelihood. Strict adherence to indications for transthoracic echocardiography and transesophageal echocardiography may help to facilitate more accurate diagnosis within the moderate likelihood category. Patients with low likelihood do not derive additional diagnostic benefit with echocardiography although other factors such as physician reassurance may continue to drive diagnostic demand.