The purpose of this study was to compare the clinical performance of in vivo vs in vitro Sv02 determinations in ICU patients using the three most current commercially available in vivo oximetry systems (Oximetrix 3, SAT-2, and HEM0PR02). Evaluation of a monitor requires that both “accuracy” and “reproducibility” are quantified. Previous studies of the performance of PA oximetry catheters have relied heavily on the use of regression analysis and correlation coefficients as a method of comparison. Clearly, each of these systems correlate with CO-oximetry and all are capable of reflecting Sv02. It has been suggested, however, that correlation coefficients constitute an inappropriate statistical tool for assessing agreement.
A high level of correlation requires only that a plot of paired data points lie along a straight line, while agreement requires that the data points lie along a line of identity. In addition, the apparent strength of a correlation can be increased by increasing the range of the values tested. Indeed, the correlation coefficients of the three systems compared with COoximetry are remarkably similar (0.85, 0.86, and 0.83 for Oximetrix 3, SAT-2, and HEM0PR02, respectively; Table 4). Alternatively, performance can be defined by bias (see Eq 1 and 3) as a measure of the constant or proportional error that may occur across a range of measurements. Likewise, reproducibility can be defined by precision (see Eq 2 and 4), which measures the nonsystematic or random variation that occurs with repeated measurements.