Directly measured skin pigmentation to mitigate pulse oximetry bias

About This Grant

ABSTRACT Bias in pulse oximetry accuracy is associated with significant delays in care and unrecognized eligibility for therapeutics among patients with darker skin pigmentation. However, very few pulse oximetry studies have incorporated direct measures of skin pigmentation, and none have incorporated objective measurements. This is a problem because the bias of modern pulse oximetry devices is believed to be related to skin pigmentation. Most studies on pulse oximetry bias are retrospective and rely on self-reported race and ethnicity. We know pulse oximetry bias contributes to worse outcomes among non-White hospitalized patients, disproportionately impacting Black and Hispanic patients. However, significant variance in skin pigmentation exists within racial and ethnic groups. Therefore, race and ethnicity are poor surrogates for skin pigmentation. Further, race and ethnicity are social constructs that do not address the root cause of pulse oximeter bias: skin pigmentation. In 2024, the US Food and Drug Administration (FDA) recommended directly measuring skin pigmentation when evaluating pulse oximeters. Accordingly, there is a critical need to prospectively determine the effects of directly measured skin pigmentation on pulse oximeter performance. To address this gap, we will test our overarching hypothesis that darker skin pigmentation will be associated with increased pulse oximetry bias. Pulse oximeter performance, as defined by the correlation between a non-invasive/continuous pulse oximeter (SpO2) and an invasive/intermittent laboratory measure of arterial oxygen saturation (SaO2), is sub-optimal among patients with darker skin pigmentation. Landmark studies demonstrated that non-White patients experience a greater incidence of hypoxemia (SaO2<88%) compared to White patients when SpO2 values are normal (i.e., unrecognized hypoxemia). Unrecognized hypoxemia increases the risk of cardiac arrest, organ dysfunction, and is associated with an up to 3-fold increase in mortality among hospitalized patients. Skin pigmentation may affect the accuracy of SpO2 measurements by introducing error from systematic bias (e.g., consistently higher SpO2 values at a given SaO2). However, retrospective studies can neither directly measure skin pigmentation nor collect arterial blood gases for paired SpO2-SaO2 comparisons. We will leverage data from our ongoing, federally funded SAVE-O2 AI multicenter clinical trial (NCT06374225), comparing a closed- loop autonomous oxygen titration device versus usual care in 300 acutely ill adult patients at three US tertiary care hospitals. We are directly measuring skin pigmentation in all patients via two validated scales plus an objective spectrophotometer. We will use data from the SAVE-O2 AI trial to determine the association between skin pigmentation and pulse oximetry bias (Aim 1). We will then create a prediction tool incorporating skin pigmentation to establish personalized SpO2 targets for hospitalized patients (Aim 2). These specific aims will generate preliminary data to support a clinical trial that will address our long-term goal of validating personalized SpO2 targets based on skin pigmentation to mitigate the risks of unrecognized hypoxemia.