Method of assessing the reproducibility of blood flow measurement: factors influencing the performance of thermodilution cardiac output computers.

Measurements of blood flow by three different makes of thermodilution cardiac output computer in an artificial circulation were analysed by linear regression against absolute flow measured by timed blood volume collection. For each computer the horizontal distance between the 95% confidence limits for a single prediction was calculated at a standard flow rate of 5 litres per minute. This measurement represents the range of flow rates that could give rise to an identical measurement and provides a summary of the reproducibility of the computer's results and its ability to detect a change of flow rate. This measurement was used to evaluate the effect on each computer's performance of pulsatile or continuous flow, injectate volume, and injectate temperature. With continuous flow the optimum results were 1.8, 0.85, and 0.85 litres per minute and with pulsatile flow they were 1.3, 1.05, and 1.65 litres per minute. There was generally a deterioration in performance when pulsatile flow was evaluated. Under the conditions of the experiment optimum performance in both flow modes was obtained with 5 ml of ice cold injectate, but these findings cannot necessarily be extrapolated to the clinical situation. With pulsatile flow the overall range of blood flows that could give rise to identical measurements were for each computer 2.0, 1.5, and 3.1 litres per minute, corresponding to 40, 30, and 62% changes of the standard flow rate of 5 litres per minute.