Evaluation of inter-observer variability regarding aortic and mitral valve findings on transesophageal echocardiograms ordered for suspected endocarditis

Background Transesophageal echocardiography (TEE) is the gold standard for the detection of valvular vegetations (VV). Differentiating small VV from degenerative changes is challenging and prone to inter-observer variability. We evaluated inter-observer agreement regarding aortic (AV) and mitral valve (MV) findings on TEEs ordered for suspected infective endocarditis (IE). Methods A total of 349 consecutive TEEs were evaluated. Studies were classified as "definite, possible, or no" IE with valve masses classified further by morphology. Nine faculty echocardiographers scored randomly selected TEEs of the AV (N = 38) and MV (N = 35). Inter-reader variability was calculated using the Fleiss/Scott Kappa (Kf). Results Positive blood cultures were present in 81% and 45% had definite IE by the modified Duke criteria. There was moderate reader agreement regarding the presence of a valvular mass for both the AV (Kf = .41, 95% CI [.30-.53]) and MV (Kf = .49, 95% CI [.34-.65]). For diagnosis of IE, there was fair agreement for the AV (Kf = .29, 95% CI [.18-.42]) and moderate agreement for the MV (Kf = .53, 95% CI [.36-.70]). Masses described as large, multi-lobulated, or pedunculated were more frequently categorized as clinical IE, (p < .006, both valves), however those with filamentous lesions were not (p < .001, both valves). Conclusions In a large academic center, the inter-observer agreement for the presence of a left sided valvular mass was moderate and agreement regarding the final diagnosis of IE was fair to moderate, with better agreement among readers evaluating the MV. Lesion morphology is associated with the clinical diagnosis of IE.