In contrast, the diagnostic agreement was good between the visually assessed LVEF using the US and TTE (Kappa: 0.75; CI 95%: 0.63 to 0.87). Similarly, the visually assessed LVEF using the US and TTE was in agreement with the measured reference values (Kappa: 0.75; 95% CI: 0.64 to 0.87 and Kappa: 0.70; 95% CI: 0.59 to 0.82, respectively). When compared with the reference LVEF, the visual assessment using the US overestimated LVEF in nine patients and underestimated LV systolic function in a single patient (Table (Table2).2). Similarly, TTE led to visually overestimate LVEF in 12 patients and to underestimate LV systolic function in 8 patients when compared to reference measurements (Table (Table3).3). When the reference LVEF was in the normal or increased range of values, LV systolic dysfunction was erroneously identified in only three patients. Regardless of the ultrasound system used, most inaccurate visual assessments of LVEF were related to an inadequate evaluation of the severity of LV systolic dysfunction or to the erroneous distinction between a normokinetic and a hyperkinetic LV (Tables (Tables22 and and33).Table 2Visual assessment of left ventricular ejection fraction (LVEF) using the ultrasound stethoscope*Table 3Visual assessment of left ventricular ejection fraction (LVEF) using the full-feature ultrasound system*DiscussionIn this study, LVEF could not be accurately predicted in ICU patients by the sole physical examination and the knowledge of a previously determined LVEF value failed to significantly improve the clinical judgement of the front-line intensivist. In contrast, the herein tested new generation US allowed an accurate semi-quantitative assessement of LVEF when compared with standard TTE, during a focused, rapid-to-perform examination.Previous studies have long shown that physical examination was inaccurate in predicting the hemodynamic status of ICU patients (for example, cardiac index, cardiac filling pressures, systemic vascular resistance) when using right heart catheterization as a reference. In these series, the cardiac index was adequately graded as low, normal or high in only approximately half of the cases when compared to measurements obtained by the thermodilution technique [1,2]. Similarly, front-line intensivists adequately predicted LVEF in only 64 of our patients (68%), as reflected by a poor diagnostic agreement between the clinical assessment and the reference LVEF value. Importantly, the use of an US as an extension of the physical examination markedly improved the clinical evaluation of cardiac function. In a systematic review of the literature, Badgett et al.