Pulse oximeter sensor accuracy does matter. We see evidence of the impact of inaccurate plus oximeter sensors in mortality reviews where health care professionals have not known which sensor to believe.

What would happen if health care professionals and patients really knew what was going on with respect to inaccurate sensors?

Would simulator manufacturers that have failed to clearly advise that their products do not test sensor accuracy be liable?

It provides a succinct, easily read and understood description of the basic principles. Unfortunately it misleads the reader by asserting that accuracy is solely a function of design, faithfully reproduced at manufacture, and that subsequent verification is neither possible, nor needed.

In practice, and in respect of sensor manufacture, this in fact can be far from the reality. In addition, there seems to be evidence (from spectral testing) that suggests an ‘ageing’effect, particularly with IR emitters over time. Any deviation from the actual wavelengths used during controlled desaturation studies to derive the original R curve, whether at manufacture, or in subsequent use, will alter that R curve, and will bias the accuracy of the pulse oximeter. A difference of just a few nanometres in the Red emitter wavelength is sufficient to cause inaccuracy beyond the standards requirement of 3% (and without this error being further compounded by biological variability).

The article acknowledges that functional testers have a role; they can be an aid to verify that any given system is functioning; they cannot verify that it is providing accurate readings, because they have no means of analysing/verifying the emitted wavelengths. Functional testing is important to ensure that monitors are working correctly (alarms etc), but this does absolutely nothing for establishing confidence in accuracy.

It is disappointing that this article has not made this distinction, and further confuses the reader by requiring a verification of accuracy of 3% in the performance test sheet (The Medical Equipment QA book linked to this article, Page 124). Whilst the requirement is correct, the functional testers described cannot verify this.

Verified functionality does not imply performance accuracy.

The publications and studies alluded to in the replies to this article have collectively sampled a large number of sensors, both new and used. The many errors identified would suggest that we should be at least as concerned with measuring sensor wavelength accuracy, as with verifying instrument function, and yet there is no acknowledgment of this as a test requirement.

Pulse oximetry at its inception in the 1980’s was hailed as “the most significant technological advance ever made in monitoring the well being of patients during anaesthesia, recovery and critical care”.
Today, we are more dependent (in the UK at least) than ever upon accuracy, since our Oxygen therapy policies are now predicated upon treating to a pulse oximeter saturation reading for diagnosis and treatment.

If the essential requirements of the relevant standards are to be met (accuracy to within 3%), and patient safety to be maintained, then accurate RED/IR wavelengths are central to achieving this, and the sooner agreement is reached upon the methods of ensuring it, the better.

Standards are often written with the benefit of hindsight available at the time they were written. Standards are the lowest common denominator that we should work to, but not be needlessly restricted by.

There appears to be some confusion over the process where the calibration data known as the R curve is established during the design phase, and the checking of accuracy of replacement sensors. The R curve is established by invasive desaturation studies where the results from a sensor of particular spectral properties are related to co-oximeter data. For replacement sensor to be accurate the emitter wavelengths must be a close match to the sensor used to establish the original R curve. As stated in the Pulse oximetry Standard (ISO 80601-2-61) – sensor accuracy is the match between the centre wavelengths emitted by the sensor and the calibration code used with that sensor.

The Fluke document referenced here is a long version of the list in the original article, and does not consider the impact and effect of replacement sensors. Sensor accuracy is important in pulse oximeter’s role in preventing adverse patient events. Users should be made aware that Healthcare Technology Management personnel following this referenced protocol do not test the correct match (accuracy) of the sensors, and that after an Annual Check of a monitor that the risk of a significant inaccuracy due to the sensor remains. Published data suggests that 30% of the sensors in use are unacceptable when tested. The cost benefit of appropriate clinical management of a patient and the cost risk of mismanagement needs to be remembered – this is a cost risk to a healthcare provider and not just the cost to an Engineering Department.

It should also be noted the IEC standard 80601-2-61 states, “There is today no accepted method of verifying the correct calibration of Pulse Oximeter Probe and Pulse Oximeter Monitor combination other than testing on human beings” (2014: line 2500).

Due to space limitations, this article does not discuss testing methods. We’d like to invite you download the full “Medical Equipment Quality Assurance” book referenced in this article to learn more about pulse oximetry test procedures and methods.

http://support.fluke.com/biomedical/Download/Asset/3276553_7150_ENG_A_W.PDF (Page 123)

The results I found were really rather sobering.
In hospital one ~3% of the sensors were classed as unacceptable (unacceptable = error > +/- 3%).
In hospital two and three over one third of the sensors tested were classed as unacceptable.

Hospital one was unusual in that all of its sensors were of one make, which seems to have a much lower proportion of inaccurate sensors.

Scarily, the sensor with the highest bias of +8% was found on a patient in theatre. This patient’s true O2 saturation could have been significantly lower than that displayed on the monitor, with the potential for hypoxic damage.
The sensor with the largest low bias (-7%) was found in an anaesthetics room within the same hospital trust.

Depending which of these sensors was applied to the patient there could be a wide variety in what the sensors may read. For example:
– A patient with a true saturation of 90% could have a SpO2 value displayed of 94% (+/-3%) or 86% (+/-3%).
– With a SpO2 display of 90% the patient could have a true saturation of 84%(+/-3%) or 92%(+/-3%)

It is rather worrying that our hospitals have such a high proportion of inaccurate sensors! We need to start testing them for accuracy in addition to functionality!

Sensor accuracy is dependent on placing LEDs of the correct wavelength in the sensor during production, and minimal changes in the spectral properties of the LEDs post production. This does not always happen and sensors can be wrong from new and can age.

The monitor is totally reliant on the red and IR wavelengths emitted by the sensor being correct. If these values are not correct the SATs values displayed on the monitor will not be correct. The SATs errors are greater at lower oxygen saturation values.

A short demonstration of a high reading sensor being tested can be viewed at: https://www.youtube.com/watch?v=pNxIAD_a_50

This error is mainly due to an error in the wavelength from the red LED. In use this sensor could lead to oxygen therapy and other important interventions being deferred or not given at all. No simulator or functional tester is capable of detecting this fault.

My concern with the article is that it does not really emphasise:
a. Limitations of simulators: that all that virtually all the spo2 simulators on the market can perform is indicating that the device is working, not that it is working accurately. Indeed, it could be argued that the simulator does little more than can be achieved by the biomed using the device on his/her own finger

b. The importance of accuracy, as other have commented on

c. That accuracy depends on the integrity of the sensor and in particular the wavelenghts of the light emmitted by the red and infrared LEDS. And that changes in the wavelenght from the design specification (poor batch, change in wavelenght with operating conditions) are likely to affect the accuracy

d. That methods are available to check the wavelenght of the LED lights sources

Best wishes

John