A Comparison of Three Pulmonary Artery Oximetry Catheters in Intensive Care Unit Patients (18)

Despite the limitations noted, there are differences in the performance of the three systems. The difference plots of in vivo vs in vitro Sv02 (Fig 1) demonstrate that both Oximetrix 3 and SAT-2 have a systematic error or inaccuracy (bias) that is consistent. The Oximetrix 3 group consistently underestimated in vitro SvOa by approximately two percentage units (bias =—1.98 percent saturation), while the SAT-2 group consistently overestimated in vitro Sv02 (bias= +1.80 percent saturation). The respective underestimation and overestimation within these two groups were constant over the range of values tested. The HEM0PR02 group also demonstrated an overall tendency to underestimate in vitro Sv02 (bias = – 2.28 percent saturation); however, this systematic error was not constant over the range of values tested. The bias demonstrated by the HEM0PR02 group appears proportional to the value tested, ranging from nearly — 5 percent saturation at low target Sv02 values (40 percent to 50 percent) to approximately — 1 percent at the higher target saturation values (80 percent to 90 percent).
More disturbing than either the relatively constant negative or positive bias exhibited by the Oximetrix 3 and SAT-2 or the nonconstant bias of the HEM0PR02 are the wide 95 percent confidence intervals relative to CO-oximetry. A constant bias, or even a nonconstant bias that is a function of the Sv02 target value, can be corrected by modifications in the computer algorithm used to calculate saturation.