The most accurate way to determine the "true" oxygen saturation level is to take a blood sample and run tests against the sample. Even so, repeated tests on the same sample may yield slightly different results [1]. Unlike using a ruler to obtain the distance between two points, there are too many factors involved. The cost associated with this test is high and can only be performed in a medical lab.

Pulse oximeter provides an indirect method. The derived blood SpO2 level is calculated from on a formula based on the light absorption by venous vessels. 1% to 2% variation from the "true" value is normal with this technology. The accuracy of a pulse oximeter is stated in terms of the root-mean-square (rms) difference between measured SpO2 values and co-oximetry (SaO2) reference values. With a accuracy rating of ±2%, a well-behaved oximeter may sometimes have some readings that are off by as much as 6%. Pulse oximeter is a sensitive device and many factors, such as random noise, motion, and physiological conditions, can cause the erroneous readings [1]. The accuracy published in the specifications of a pulse oximeter is based on the performance of a collection of test samples rather than each individual reading.

In general, with an accuracy rating of ±2%, saturation readings in the 90-100% range would have an accuracy of less than 2%; saturation readings in the 80-90% range would have an accuracy of about 2%; saturation readings in the 70-80% range would have an accuracy of greater than 2%. The accuracy would also decrease with low blood perfusion and low heart rate.

References

1. Paul Batchelder and Dena Raley, "Maximizing the Laboratory Setting for Testing Devices and Understanding Statistical Output in Pulse Oximeter", International Anesthesia Research Society, Vol. 105, No. 8, August 2007.