ACCURACY AND PRECISION
Numerous studies have compared CO measurements from the ODM with those from the pulmonary artery catheter (PAC), with good overall agreement (see  for systematic review). However, there are limitations in comparing new technologies to the PAC. Despite being deemed the clinical ‘gold standard’ in CO measurement, the PAC has an error of ±20% .
As a result this makes it difficult to understand the true accuracy of newer technologies that are calibrated against and/or compared to the PAC. Because of these limitations in the accurate measurement of CO, the precision or reproducibility of the technology becomes important.
The precision of a technology can be assessed by taking repeated measurements of SV or CO in a haemodynamically stable patient over a short period or time. The variability of the data set can then be calculated. This is most commonly reported as the coefficient of variation (the standard deviation (SD) as a percentage of the mean value).
The figure below shows the clustering of normally distributed data around the mean value.
~70% of data points will fall within 1 SD of the mean
~95% of data points will fall within 2 SD of the mean
~99% of data points will fall within 2.5 SD of the mean
This normal distribution of data enables the determination of confidence intervals. In CO monitoring terms this is the change in measured CO (or SV) required to be x% confident that the change is real and not due to inherent measurement error within the technology. Based on the above…
If the SV/CO changes by 1 SD, the user can be ~70% sure that the change is real.
If the SV/CO changes by 2 SD, the user can be ~95% sure that the change is real.
If the SV/CO changes by 2.5 SD, the user can be ~99% sure that the change is real.
In order to measure whether a variable has changed, the ‘amount’ of measured change must be greater than the precision of the technology. If the amount of change is 2.5 times the precision, the user can be 99% sure they have measured a real change in the variable. But, what are the values of ‘1 SD’ (or coefficient of variation) and therefore the 99% confidence intervals for each of the different CO measurement technologies…i.e., how precise are they?
The table below summarises the available published precision data on the different CO monitoring technologies. At the bottom of the table, the ‘mean’ is the mean coefficient of variation, and the 99% CI is the change in SV/CO required in response to an intervention to be 99% sure that the change is real and not due to measurement error (this is often called the ‘least significant change’).
PAC, pulmonary artery catheter; ODM, oesophageal Doppler monitor; PPWA, pulse pressure waveform analysis.
The precision of a technology dictates its ability to guide fluid management. The 10% SV optimisation algorithm used to optimise SV is specific to the oesophageal Doppler device, and is evidence-based. As outlined above, the oesophageal Doppler can precisely identify a 10% change in SV. Other technologies that are less precise may not be as effective in guiding fluid management based on this algorithm. Furthermore, the precision of the technology is the likely reason for the numerous positive clinical outcome studies (and the subsequent NICE guidance ), which support the use of oesophageal Doppler for fluid management.
- Dark, P. M. and M. Singer, The validity of trans-esophageal Doppler ultrasonography as a measure of cardiac output in critically ill adults. Intensive Care Med, 2004. 30(11): p. 2060-6.
- Yang, X.X., L.A. Critchley, and G.M. Joynt, Determination of the precision error of the pulmonary artery thermodilution catheter using an in vitro continuous flow test rig. Anesth Analg, 2011. 112(1): p. 70-7.
- NHS: National Institute for Health and Clinical Excellence. CardioQ-ODM Oesophageal Doppler Monitor. http://guidance.nice.org.uk/MTG3, 2011