An assessment of the accuracy of pulse oximeters

Transcription

1 Anaesthesia 212, 67, Original Article doi:1.1111/j x An assessment of the accuracy of pulse oximeters Q. J. W. Milner 1 and G. R. Mathews 2 1 Consultant, Department of Anaesthesia, Royal Devon & Exeter Foundation Trust, Exeter, UK 2 Director, The Electrode Company, Monmouthshire, UK Summary Peripheral pulse oximetry has become a core monitoring modality in most fields of medicine. Pulse oximeters are used ubiquitously in operating theatres, hospital wards, outpatient clinics and general practice surgeries. This study used a portable spectrometer (Lightman Ò, The Electrode Co. Ltd., Monmouthshire, UK) to measure the emission spectra of the two light emitting diodes within the pulse oximeter sensor and to determine the accuracy of 847 pulse oximeters currently in use in 29 NHS hospitals in the UK. The standard manufacturing claim of accuracy for pulse oximeters is ± 2 3% over the range of 7 1% S p O 2. Eighty-nine sensors (1.5%) were found to have a functional error of their electrical circuitry that could cause inaccuracy of measurement. Of the remaining 758 sensors, 169 (22.3%) were found to have emission spectra different from the manufacturers specification that would cause an inaccuracy in saturation estimation of > 4% in the range of 7 1% saturation. This study has demonstrated that a significant proportion of pulse oximeter sensors may be inaccurate.... Correspondence to: Dr Q. J. W. Milner quentin.milner@rdeft.nhs.uk Accepted: 23 November 211 Peripheral pulse oximetry is a core monitoring modality in most fields of medicine. A number of treatment protocols are now guided in part by the results of peripheral oxyhaemoglobin saturation (S p O 2 ) measurement [1, 2]. Patient factors causing errors in the measurement of oxyhaemoglobin saturation are well known [3, 4]. Potential errors in the function of the pulse oximeter itself are less clear. Pulse oximeters use two light emitting diodes (LEDs) of specific and differing wavelength (typically 66 and 94 nm) to measure the combined absorption by a mixture of oxyhaemoglobin and deoxyhaemoglobin of red and infrared light measured using a photodiode. The photodiode measures the variation in the intensity of light falling upon it, and converts this into an electrical voltage. The ratio of the absorption at these two wavelengths is called the R value, and is compared with R values that are calibrated against direct measurements of arterial oxyhaemoglobin saturation (S a O 2 ) and arterial partial pressure of oxygen (P a O 2 ) for an individual model of pulse oximeter, using a volunteer population sample. Volunteers breathe controlled hypoxic gas mixtures to create a range of S a O 2 values between 7% and 1% against which the S p O 2 of an individual pulse oximeter sensor may be calibrated. Once in use, there is little evidence that pulse oximeters are ever re-calibrated or have their accuracy assessed. A high reliance is placed upon the continuing reliability of the wavelength of the light emitted from the LEDs. Should the wavelength of the light emitted by one or both LEDs alter, the degree of absorption will alter and the R value will change. The mathematical algo- 396 Anaesthesia ª 212 The Association of Anaesthetists of Great Britain and Ireland

2 Milner and Mathews Accuracy of pulse oximeters Anaesthesia 212, 67, rithm contained within the pulse oximeter software for the estimation of S p O 2 will then be fed incorrect information, and an inaccurate estimation of S p O 2 will be displayed on the monitor. This may in turn lead to incorrect diagnosis and treatment of a patient. The common manufacturing literature claim for accuracy for pulse oximeters is ± 2 3% S p O 2 over a declared range of 7 1%, and the International Standard for pulse oximeter manufacture ISO requires an accuracy of 4% error in this range [5]. The Lightman Ò is a portable microspectrometer manufactured by The Electrode Company (The Electrode Co. Ltd., Monmouthshire, UK) that uses an internal diode-array spectrometer to measures the exact emission spectra of the LEDs in pulse oximeter sensors (Fig. 1). Once the LED wavelengths have been measured, an internal algorithm is used to compare the observed R value with the expected R value, and the degree of bias at a range of S p O 2 values (97%, 9%, 8%, Figure 1 Lightman Ò device with a pulse oximeter sensor attached to the photodetector finger and cable plugged into a model-specific adapter. 7%) is calculated. As the absorption curve for oxyhaemoglobin is steeper at lower oxygen saturations, the degree of bias is non-linear and increases as S p O 2 decreases. Measurement with the Lightman is simple, can be performed anywhere, and each test takes < 3 s. We undertook an assessment of the accuracy of a random selection of pulse oximeters available for clinical use in UK hospitals. Methods Pulse oximetry sensors available for immediate clinical use in 29 NHS hospitals in the UK were subjected to analysis for accuracy using the Lightman device. The hospitals chosen were those that had responded to a mailshot campaign by the manufacture of the Lightman device. All the sensors included in the study were available for immediate clinical use at the time of the study. The clinical areas that were studied were those with which the local clinician or technical staff involved in the data collection were familiar; these included operating theatres, anaesthetic rooms, recovery wards, intensive care units and ward areas. The Lightman device has been programmed to identify and test specific sensors compatible with all the leading brands of pulse oximeter available in the UK, including original manufacturer and third party manufactured sensors. Calibration of the spectrometer in the Lightman device is carried out automatically and immediately before each test by use of an internal neon argon emission spectra device. A commercially available neon argon bulb reliably emits peaks of light at , and nm, spanning the wavelengths used in pulse oximetry. The pulse oximeter cable is plugged into the Lightman via a range of manufacturer specific adaptors enabling it to identify the specific model of sensor. The sensor itself is placed over a translucent window in a light diffusing photodetector finger to collect the LED emissions and transport them via a fibreoptic cable to the spectrometer. Appropriate software obtained by either reverse engineering or manufacturer co-operation is included for most pulse oximeters in use in the UK, and the Lightman identifies each specific unit once connected. The device initially determines the continuity of the electrical circuitry and brightness of the LED light emissions. The spectrometer also detects single LEDs that emit more than one Anaesthesia ª 212 The Association of Anaesthetists of Great Britain and Ireland 397

3 Anaesthesia 212, 67, Milner and Mathews Accuracy of pulse oximeters wavelength of light, known as secondary emissions. Sensors that fail on any of these parameters do not proceed to the specific spectrometry analysis and were recorded as being nonfunctional. The spectrometer within the Lightman then measures the specific red and infrared wavelengths emitted and calculates an R value for this combination. The predicted degree of bias (both positive and negative) that the emitted wavelengths will produce at S p O 2 97%, 9%, 8% and 7% is displayed. It is possible that errors in emission of both red and infrared wavelengths will combine to cancel each other out and still create a correct R value. Where this happens, the Lightman will record the sensor as behaving accurately despite incorrect function. All recordings were made by a member of the local clinical or technical staff, usually accompanied by a representative from the manufacturer of the Lightman device. All data were recorded on data collection sheets and their veracity was confirmed by signature by the supervising clinical or technical staff. Data collection included the location of clinical area, sensor manufacturer and date of study. Permission for group publication of data was obtained from the individual clinician or the engineering department manager as appropriate, and where a faulty sensor was detected, the local management of this sensor was left to the discretion of the local clinician. As no patient contact occurred in this study, ethical approval was not sought from individual hospitals. Results A total of 847 individual pulse oximeter sensors used in 29 NHS hospitals in the UK were examined using the Lightman device. The sensors were manufactured by 17 different companies (range sensors), but 1 manufacturers contributed < 1 sensors each, and 88.5% of the sensors were supplied by five companies (range ). Eighty-nine sensors (1.5%) were found to have a functional error of electrical circuitry that could cause inaccuracy of measurement. These sensors were not subjected to further spectrograph analysis. Of the remaining 758 sensors, 169 (22.3%) were found to have emission spectra different from the manufacturers specification that would cause inaccuracy of S p O 2 estimation > 4% in the 7 1% range of S p O 2 (Fig. 2). In the 8 1% range, the number of sensors producing an error < 4% was 7 (9.2%). Thus, a total of 258 of the 847 pulse oximeters (3.5%) were found not to be working according to the manufacturers specification. In none of the 29 hospitals did we find a reliable system for the periodic or cyclical calibration of pulse oximeter sensors. Discussion Milliken first introduced the term oximeter in 1942 [6], and pulse oximetry was described by Aoyagi in 1974 [7]. Since its introduction into clinical use in the 198s, pulse oximeters have become ubiquitous clinical monitoring devices used throughout medical practice from operating theatres and intensive care units to outpatient departments and general practice clinics. Cheap pulse oximeters are also widely available for purchase by the general public. Knowledge of the function of pulse oximetry and its limitations will be variable across such a large range of users [8]. The results of bedside pulse oximetry are at the heart of a wide range of clinical practice guidelines for patient care from the neonatal diagnosis of congenital heart disease to the management of chronic obstructive pulmonary disease and emergency oxygen therapy [1, 9 11]. Various brands of functional tester are available for pulse oximeter sensors, the majority using an optomechanical finger containing a light detector and an LED light emitter to deliver light to the sensor detector. These devices solely test the continuity of the circuitry within the device and do not assess accuracy. The International Organization for Standardization regulates basic safety and essential performance of pulse oximeter sensors in ISO :211(E), and recognises that the current pulse oximetry functional testers do not and cannot calibrate or measure accuracy [5]. The Lightman is unique in its ability to measure the specific wavelength of pulse oximeter LED light emissions, and thus determine the degree of bias to which a machine will be subject. Although the Lightman is not a medical device, the design, development, verification and validation of the Lightman has been carried out according to Medical Device Standards, and the technical files are audited to ISO91, ISO and the Canadian Medical Devices Conformity Assessment Scheme 398 Anaesthesia ª 212 The Association of Anaesthetists of Great Britain and Ireland

4 Milner and Mathews Accuracy of pulse oximeters Anaesthesia 212, 67, (a) 45 (b) (c) 45 (d) Figure 2 The bias of 758 pulse oximeter sensors at differeing S p o 2 values: a 97%; b 9%; c 8%; and d 7%. (CMDCAS) by an independent Notified Body. Manufacturing validation of the Lightman was conducted by performing simultaneous, sequentially hypoxic, breathe down tests using accurate and inaccurate sensors whose optical properties had been measured using a Bentham bench spectrometer, and conform to ISO :211(E). The ability of the Lightman to predict S p O 2 values accurately from LED emission data has recently been shown by Dugani et al., who compared pulse oximeter data from accurate and inaccurate sensors with those obtained from co-oximetry blood gas analysis in 63 sets of measurements from patients receiving level 1 3 critical care [12]. The present study has demonstrated that a significant proportion of pulse oximeters in daily clinical use in NHS hospitals in the UK is inaccurate and will produce misleading results. As a result of the steeper absorption curve for deoxyhaemoglobin than oxyhaemoglobin, an error of just 4 nm in emission wavelength of red light using an LED will produce an error of S p O 2 estimation of 2% at 97% saturation, 4% at 9% saturation and 7% at 8% saturation. The impact on tissue oxygenation is even greater at lower S p O 2 levels because of the sigmoid shape of the oxy-haemoglobin dissociation curve. Manufacturing specifications describe an accuracy of ± 2 3% across the S p O 2 range 7 1%. As a result of the increasing steepness of the oxy-haemoglobin dissociation curve as S p O 2 falls, the number of sensors with a bias greater than ± 4% increases between 8% and 7%, a range below most clinical situations. We therefore reported our results across both the 7 1% range, and the more clinically relevant 8 1% range. Anaesthesia ª 212 The Association of Anaesthetists of Great Britain and Ireland 399

5 Anaesthesia 212, 67, Milner and Mathews Accuracy of pulse oximeters The causes of pulse oximeter inaccuracy are broadly related to either mechanical electrical integrity or emission spectra inaccuracy. In this study, 1.5% of the sensors had detectable faults in their electrical integrity. Where there is damage to the sensor that results in electrical noise being fed back to the monitor, the monitor may measure this as a pulsatile ratio of 1:1 which will then be interpreted as a S p O 2 value of 83%. When undetected, these cable faults frequently cause the monitor to display S p O 2 readings in the low 8s regardless of the patient s true S p O 2 value. These types of cable faults are not identified by the monitor, simulators or standard electrical tests. Degradation in performance of LEDs can occur over time [13]. Simplistically, LEDs are a stack of semi-conductor metal crystals (typically aluminium, gallium and arsenic) that emit radiation when a current is passed through them (electroluminescence). Energised electrons move to higher energy levels and as they fall back, optical radiation is emitted. Faults within the structure of the semi-conductor crystals can originate at manufacture or occur over time, changing the spectral properties of the LED [14 16]. It is likely that manufacturing inaccuracy, translating laboratory specification into mass production, is another source of wavelength variability. For example, excess solder used to attach an LED to a circuit board can act as a heat sink and alter the temperature and thus optical properties of the LED. Re-usable pulse oximeter sensors, intended for multiple patient usage over a prolonged period, are also likely to suffer physical damage that may alter their optical performance This investigation examined only a small proportion of the pulse oximeters in use in each hospital. Sensors were chosen for assessment solely because of their availability within a particular clinical area and no attempt at systematic assessment of all available sensors was performed. All the tested sensors were, however, available for immediate clinical use and were tested as found. It is possible that cleaning the sensors may in some instances have improved their performance. However, as this is not routinely carried out before clinical use, it was not carried out before testing with the Lightman. With the exception of sensors still within their original packaging, it was not possible to ascertain the age or duration of use of the pulse oximeter sensors. Some sensors are left continuously powered on in a clinical environment, whereas others are only powered when used. Access to individual hospitals was organised by representatives of The Electrode Company following a response to their commercial advertising by either a clinician or hospital biomedical electrical department, and this may have inflated the number of inaccurate sensors found. When a faulty sensor was detected, further investigation and management was left to the local clinical or technical staff. This included disposal, return to original manufacturer or referral to the Medicines and Healthcare Regulatory Agency. Pulse oximeters are ubiquitous in clinical practice, and their results widely used to influence patient care. This study has used a portable microspectrometer (Lightman) to demonstrate that a significant proportion of devices in clinical use are inaccurate. All monitoring devices should undergo periodic calibration and accuracy assessment. This investigation could find no evidence that such practice occurs. Healthcare providers using pulse oximetry should ensure that these devices function correctly. Competing interests GM is a director of The Electrode Company, the manufacturer of the Lightman device. QM has no financial interest in The Electrode Company or the Lightman device. References 1. O Driscoll BR, Howard LS, Davison AG. British Thoracic Society guideline for emergency oxygen use in adults. Thorax 8; 63: Castillo A, Sola A, Baquero H, et al. Pulse oxygen saturation levels and arterial oxygen tension values in newborns receiving oxygen therapy in the neonatal intensive care unit: is 85% to 93% an acceptable range? Pediatrics 8; 121: Van de Louw A, Cracco C, Cerf C, et al. Accuracy of pulse oximetry in the intensive care unit. Intensive Care Medicine 1; 27: Wilson BJ, Cowan HJ, Zuege DJ, Zygun DA. The accuracy of pulse oximetry in emergency department patients with severe sepsis and septic shock: a retrospective cohort study. BMC Emergency Medicine 21; 1: International Organization for Standardization (ISO). Particular requirements for basic safety and essential performance of pulse oximeter equipment. ISO : Millikan GA. The oximeter: an instrument for measuring continuously oxygen-saturation of arterial blood in man. Review of Scientific Instruments 1942; 13: Yoshiya I, Shimada Y, Tanaka K. Spectrophotometric monitoring of arterial oxygen in the fingertip. Medical and Biological Engineering and Computing 198; 18: Anaesthesia ª 212 The Association of Anaesthetists of Great Britain and Ireland

6 Milner and Mathews Accuracy of pulse oximeters Anaesthesia 212, 67, Stoneham MD, Saville GM, Wilson IH. Knowledge about pulse oximetry among medical and nursing staff. Lancet 1994; 344: Mahle WT, Newburger JW, Matherne G, et al. Role of pulse oximetry in examining newborns for congenital heart disease: a scientific statement from the American Heart Association and American Academy of Pediatrics. Circulation 9; 12: Murphy R, Driscoll P, O Driscoll R. Emergency oxygen therapy for the COPD patient. Emergency Medical Journal 1; 18: Gooptu B, Ward L, Ansari SO, Eraut CD, Law D, Davison AG. Oxygen alert cards and controlled oxygen: preventing emergency admissions at risk of hypercapnic acidosis receiving high inspired oxygen concentrations in ambulances and A&E departments. Emergency Medical Journal 6; 23: Dugani S, Hodzovic I, Sindhakar S, et al. Evaluation of a pulse oximeter sensor tester. Journal of Clinical Monitoring and Computing 211; 25: Kuo LH, Kimura K, Miwa S, Yasuda T, Yao T. Role of interface chemistry and growing surface stoichiometry on the generation of stacking faults in ZnSe GaAs. Journal of Electronic Materials 1997; 26: Fukuda M. Laser and LED reliability update. Journal of Lightwave Technology 1988; 6: Chun-Chin T, Ming-Hung C, Yi-Chung H, et al. Decay mechanisms of radiation pattern and optical spectrum of high-power LED modules in aging test. IEEE Journal of Selected Topics in Quantum Electronics 9; 15: Kock JP, Reynolds KJ, Tarassenko L, Moyle JTB. The effect of varying LED intensity on pulse oximeter accuracy. Journal of Medical Engineering and Technology 1991; 15: Anaesthesia ª 212 The Association of Anaesthetists of Great Britain and Ireland 41