Potential errors in pulse oximetry. I. Pulse oximeter evaluation.

There is no absolute reference for oxygen saturation, although multiwavelength in vitro oximeters are accepted as the 'gold standard'. Regardless of whether fractional or functional saturation is used by manufacturers to calibrate their oximeters, evaluation against fractional saturation is recommended since this is the clinically relevant variable. The use of standard notation and comparisons based on bias and precision is recommended. The accuracy of pulse oximetry is intrinsically limited by the use of only two wavelengths, and is dependent on the initial calibration population. The empirical algorithms used to convert the signal to its 'readout value' and the quality control of hardware may both be important sources of variability between oximeters. Change in blood temperature may introduce errors in pulse oximeter and in vitro oximeter saturation readings, but these will be clinically insignificant. Changes in blood pH should not decrease pulse oximetry accuracy.