Cardiac output was measured 120 times in 27 critically ill patients using the thermodilution and transthoracic electrical impedance methods. Both the minimum and mean values for the distance between the inner electrodes, and a variety of values for the resistivity of blood (rho) were substituted in the Kubicek's empiric formula for calculating cardiac output by transthoracic electrical impedance. Using the mean distance between the inner electrodes and a rho-value of 150 ohm X cm gave the best agreement between the methods (mean difference 0.17 +/- 2.4 L/min). Ventilation alone or with positive end-expiratory pressure did not significantly affect the bias of the estimate, but both affected its precision when compared with measurements in spontaneously breathing patients (SD of mean difference 2.4 and 3.2 L/min, respectively, vs. 1.5 L/min). The pulmonary artery wedge pressure was significantly higher in patients with an abnormal diastolic impedance waveform (zero-wave), but there was no relationship between wedge pressure and base impedance per unit length between the measuring electrodes.