Experimental and predicted dual oximetry variability.

Abstract

We wished to determine whether the individual bias (mean difference) and precision (standard deviation of the difference) values of 2 variables, arterial oxygen saturation (SaO2) and mixed venous oxygen saturation (SvO2), could be used to predict the bias and precision values of the combined dual oximetry variable (SaO2-SvO2). We simultaneously measured SaO2 by pulse oximetry and arterial blood gas co-oximetry and SvO2 by fiberoptic reflectance oximetry pulmonary artery catheter and venous blood gas co-oximetry in 238 data sets from 55 patients. Three different methods were used to predict the standard deviation of the difference of (SaO2-SvO2) [s delta(SaO2-SvO2)]: simple sum, root mean square (RMS) error, and RMS error with correction term. We derived the equation for the RMS error with correction term because initial results showed that the simple sum and RMS error methods did not predict s delta(SaO2-SvO2) well. The correction term accounts for the non-independence of simultaneous SaO2 and SvO2 measurements. The observed overall bias of the SaO2, SvO2, and (SaO2-SvO2) measurement methods were 0.17, -1.76, and 1.94, respectively. The observed overall s delta(SaO2-SvO2) of the (SaO2-SvO2) measurement method was 5.12. The simple sum method overestimated the actual s delta(SaO2-SvO2) by 38%, the RMS error method differed from the actual s delta(SaO2-SvO2) by 3%, and the RMS error with correction term method matched the actual s delta(SaO2-SvO2). The bias of a (SaO2-SvO2) measurement method is simply the bias of the SaO2 measurement method less the bias of the SvO2 measurement method. s delta(SaO2-SvO2) is best predicted by the derived equation, RMS error with correction term. The same principles and equations also apply to other situations in which 2 variables with the same dimensions are combined into 1 variable, such as (PaCO2-EtCO2) gradients and perfusion-pressure gradients. Although the difference between the s delta(SaO2-SvO2) predicted by the RMS error equation and the derived RMS error equation with correction term was small, the difference may be significant for other combined variables.