Cardiac output measurement by the thermodilution method: an in vitro test of accuracy of three commercially available automatic cardiac output computers

Abstract

Objective: To describe the accuracy and the reproducibility of the thermodilution flow measurements obtained using 3 commercially available cardiac output computers commonly used in intensive care units.

Design: An experimental "in vitro" study. Twelve different values of control flow (Qctr) were measured (Qmsr) using 3 different cardiac output computers (Abbott Critical Care System, Oximetrix 3 SvO2/CO Computer, Baxter Oximeter/Cardiac Output Computer SAT-1; American Edwards Laboratories, 9520 A Cardiac Output Computer). Standard equipment and techniques were employed, taking account of the specific weight and heat of warm water relative to blood. In addition, separate sets of measurements were performed in order to investigate the effect on Qmsr of some variables which may influence the "indicator" loss (time for injection, depth of immersion of the catheter, temperature of the injected fluid).

Setting: Our laboratory, inside the intensive care unit.

Measurements and results: The analysis of the linear regression of Qmsr versus Qctr (r values between 0.992 and 0.984; residual standard deviation values comprised between 0.24 and 0.49 l/min; intercepts and slopes not significantly different from identity line), the values of the percentage errors (PE = [Qctr-Qmsr].100/Qctr; PE mean values 7.9, 5.0 and 13.1), and those of the coefficients of variability (CV = standard deviation mean value, %; CV mean values 5.4, 5.8 and 4.6), show a good level of accuracy and reproducibility of the measurements. Our data confirm previously reported results. Furthermore, the cumulative effect of variables capable of influencing the "indicator" loss, even if corrected according to the "calculation constant" the manufacturers provide, was found to result in statistically significant changes of Qmsr.

Conclusion: The accuracy and reproducibility of the automatic cardiac computers tested is sufficient for practical clinical purpose. It may also depend on the modality of injection of the cooling bolus, which may significantly influence the effective "indicator" losses.