Echocardiographic agreement in the diagnostic evaluation for infective ă endocarditis

International audienceEchocardiography is essential for the diagnosis and management of ă infective endocarditis (IE). However, the reproducibility for the ă echocardiographic assessment of variables relevant to IE is unknown. ă Objectives of this study were: (1) To define the reproducibility for IE ă echocardiographic variables and (2) to describe a methodology for ă assessing quality in an observational cohort containing site-interpreted ă data. IE reproducibility was assessed on a subset of echocardiograms ă from subjects enrolled in the International Collaboration on ă Endocarditis registry. Specific echocardiographic case report forms were ă used. Intra-observer agreement was assessed from six site readers on ten ă randomly selected echocardiograms. Inter-observer agreement between ă sites and an echocardiography core laboratory was assessed on a separate ă random sample of 110 echocardiograms. Agreement was determined using ă intraclass correlation (ICC), coverage probability (CP), and limits of ă agreement for continuous variables and kappa statistics ă (kappa(weighted)) and CP for categorical variables. Intra-observer ă agreement for LVEF was excellent [ICC = 0.93 +/- 0.1 and all pairwise ă differences for LVEF (CP) were within 10 %]. For IE categorical ă echocardiographic variables, intra-observer agreement was best for ă aortic abscess (kappa(weighted) = 1.0, CP = 1.0 for all readers). ă Highest inter-observer agreement for IE categorical echocardiographic ă variables was obtained for vegetation location (kappa(weighted) = 0.95; ă 95 % CI 0.92-0.99) and lowest agreement was found for vegetation ă mobility (kappa(weighted) = 0.69; 95 % CI 0.62-0.86). Moderate to ă excellent intra- and inter-observer agreement is observed for ă echocardiographic variables in the diagnostic assessment of IE. A ă pragmatic approach for determining echocardiographic data ă reproducibility in a large, multicentre, site interpreted observational ă cohort is feasible