Neonatal Oximeter Bias Study

Overview

The study aims to determine if there is a systematic difference (bias) between pulse oximeters used in control systems for automated oxygen delivery (A-FiO2) and those used in monitoring systems. When using A-FiO2 systems there are commonly two oximeter probes on the infant with two difference readouts. Nurses report frustration that the two readings are often markedly different. It is understandable that physiological differences between sensor sites might reflect different regional oxygen saturation levels. It is also possible that there is a relevant systematic bias between difference monitors and sensors. Therefore, a large systematic multicenter study is needed to determine whether these frequent differences should be ignored as physiological noise or considered clinically relevant.

The accuracy of clinical pulse oximeters is specified as 3% RMS (root mean squared of the bias and variance) compared to arterial SaO2, in the range of 70-100% SpO2. This combines errors from bias and from scatter (imprecision). These validations are carried out under ideal laboratory conditions; that is, on healthy volunteers who undergo experimental desaturations, and not in the routine clinical environment. In a large multicenter clinical observation study, Ross et al in 2013 identified important related concerns. Specifically, they found that many of the clinical measurements were outside the 3% accuracy envelop. Specifically, they reported that bias varied depending on the level of saturation, the oximeter brand, sensor, race and site perfusion. Others have identified relevant differences in oximeters, oximeter sensors \[Maiwald\], and skin pigmentation. Our study would be the first study in the neonatal ICU evaluating bias to consider different oximeters, sensors and sensor sites. It is highly topical in that changes in average SpO2 of 3% have been associated with excess neonatal mortality and morbidity. This study is planned to investigate the source and magnitude of differences in SpO2 readings from oximeter sensors at different sensor sites that are routinely noticed by clinicians. While these differences are most often ignored and attributed to sensor site perfusion, if systematic they could have marked impact on mortality and morbidity. The aim of the investigation is to provide practical guidance relating to SpO2 bias among oximeters, oximeter sensors and sensor location. While taking advantage of the clinical need for 2 oximeter sensors, the results will be applicable to all neonatal oximeter monitoring. An observational design was selected to take place in centers using automated FiO2 control systems (A-FiO2) that routinely require the use of one sensor for control and another for monitoring. This is a refinement of the approach used in the often-cited multicenter evaluation of neonatal SpO2 exposure. Thus, without an investigational intervention the observational design is well suited to collect comparable data from multiple centers.