The study's goal was to determine if cardiac output (CO), obtained by impedance cardiography (ICG), would be improved by a new equation N, implementing a square root transformation for dZ/dtmax/Z0, and a variable magnitude, mass-based volume conductor Vc. Pulmonary artery catheterisation was performed on 106 cardiac surgery patients pre-operatively. Post-operatively, thermodilution cardiac output (TDCO) was simultaneously compared with ICG CO. dZ/dtmax/Z0 and Z0 were obtained from a proprietary bioimpedance device. The impedance variables, in addition to left ventricular ejection time TLVE and patient height and weight, were input using four stroke volume (SV) equations: Kubicek (K), Sramek (S), Sramek-Bernstein (SB), and a new equation N. CO was calculated as SV x heart rate. Data are presented as mean +/- SD. One way repeated measures of ANOVA followed by the Tukey test were used for inter-group comparisons. Bland-Altman methods were used to assess bias, precision and limits of agreement. P< 0.05 was considered statistically significant. CO implementing N (6.06 +/- 1.48 l min(-1)) was not different from TDCO (5.97 +/- 1.41 l min(-1)). By contrast, CO calculated using K (3.70 +/- 1.53 l min(-1)), S (4.16 +/- 1.83 l min(-1)) and SB (4.37 +/- 1.82 l min(-1)) was significantly less than TDCO. Bland-Altman analysis showed poor agreement between TDCO and K, S and SB, but not between TDCO and N. Compared with TDCO, equation N, using a square-root transformation for dZ/dtmax/Z0, and a mass-based Vc, was superior to existing transthoracic impedance techniques for SV and CO determination.