Analysis of Discrepancies Between Pulse Oximetry and Arterial Oxygen Saturation Measurements by Race and Ethnicity and Association With Organ Dysfunction and Mortality

Wong, An-Kwok Ian; Charpignon, Marie-Laure; Kim, Han; Josef, Christopher S.; Hond, Anne de; Fojas, Jhalique Jane R.; Tabaie, Azade; Liu, Xiaoli; Mireles-Cabodevila, Eduardo; Carvalho, Leandro; Kamaleswaran, Rishikesan; Madushani, R. W. M. A.; Adhikari, Lasith; Holder, Andre; Steyerberg, Ewout W.; Buchman, Timothy G.; Lough, Mary E.; Celi, Leo Anthony

doi:10.1001/jamanetworkopen.2021.31674

Cited by 148 publications

(171 citation statements)

References 26 publications

Supporting

Mentioning

143

Contrasting

Order By: Relevance

“… 20 However, the need for training ANNs to the full range of Fitzpatrick skin types I to VI is also indicated by recent observations in pulse oximetry, where darker skin results in a small but larger number of undetected hypoxemia events. 37 Furthermore, all measurements were performed on the volar forearm during occlusion provocations. Our results suggest that a training data set of 19 subjects is enough for training an

algorithm that can estimate

in low-pigmented forearm skin.…”

Section: Discussionmentioning

confidence: 99%

Multispectral snapshot imaging of skin microcirculatory hemoglobin oxygen saturation using artificial neural networks trained on in vivo data

et al. 2022

View full text Add to dashboard Cite

. Significance: Developing algorithms for estimating blood oxygenation from snapshot multispectral imaging (MSI) data is challenging due to the complexity of sensor characteristics and photon transport modeling in tissue. We circumvent this using a method where artificial neural networks (ANNs) are trained on in vivo MSI data with target values from a point-measuring reference method. Aim: To develop and evaluate a methodology where a snapshot filter mosaic camera is utilized for imaging skin hemoglobin oxygen saturation ( ), using ANNs. Approach: MSI data were acquired during occlusion provocations. ANNs were trained to estimate with MSI data as input, targeting data from a validated probe-based reference system. Performance of ANNs with different properties and training data sets was compared. Results: The method enables spatially resolved estimation of skin tissue . Results are comparable to those acquired using a Monte-Carlo-based approach when relevant training data are used. Conclusions: Training an ANN on in vivo MSI data covering a wide range of target values acquired during an occlusion protocol enable real-time estimation of maps. Data from the probe-based reference system can be used as target despite differences in sampling depth and measurement position.

show abstract

algorithm that can estimate

in low-pigmented forearm skin.…”

Section: Discussionmentioning

confidence: 99%

Multispectral snapshot imaging of skin microcirculatory hemoglobin oxygen saturation using artificial neural networks trained on in vivo data

et al. 2022

View full text Add to dashboard Cite

show abstract

“…Underestimated SpO 2 readings also have the potential to be harmful, resulting in unnecessary treatment with oxygen (and the risk of hyperoxaemia) and wider impacts such as delayed hospital discharge. Two recent diagnostic studies provide evidence on clinical implications resulting from the bias in pulse oximetry for blood oxygen saturation levels[7,69]. In these studies people described as Black had a higher risk of ‘ occult hypoxemia that was not detected by pulse oximetry ’ compared with those described as White.…”

Section: Discussionmentioning

confidence: 99%

“…Firstly, seven studies considering the impact of skin pigmentation and eight studies considering ethnicity compared SpO 2 -SaO 2 bias data between different subgroups and presented only tests of significance results, rather than SpO 2 and SaO 2 data per se at each subgroup level. Also, at least two studies used diagnostic accuracy design that only presented proportions of participants with specific ranges of SpO 2 in relation to specific SaO 2 values, again rather than SpO 2 and SaO 2 data per se [7,69]. We contacted authors of these studies to request relevant data and received data for two studies.…”

Section: Discussionmentioning

confidence: 99%

The effects of skin pigmentation on the accuracy of pulse oximetry in measuring oxygen saturation: a systematic review and meta-analysis

Shi

Goodall

Dumville

et al. 2022

Preprint

View full text Add to dashboard Cite

Background Pulse oximetry was widely used in hospitals and at home to monitor blood oxygen during the COVID-19 pandemic. There have been concerns regarding potential bias in pulse oximetry measurements for people with dark skin. We aimed to assess the effects of skin pigmentation on the accuracy of oxygen saturation measurement by pulse oximetry (SpO2) compared with the gold standard SaO2 measured by CO-oximetry. Methods We searched Ovid MEDLINE, Ovid Embase, and EBSCO CINAHL Plus (up to December 2021), as well as ClinicalTrials.gov and World Health Organization International Clinical Trials Registry Platform (up to August 2021). We identified studies comparing SpO2 values in any population, in any care setting, using any type of pulse oximeter, with SaO2 by standard CO-oximetry; and measuring the impact of skin pigmentation or ethnicity on pulse oximetry accuracy. We performed meta-analyses for mean bias (the primary outcome in this review) and its standard deviations (SDs) across studies included for each subgroup of level of skin pigmentation and ethnicity. We calculated accuracy root-mean-square (Arms) and 95% limits of agreement based on pooled mean bias and pooled SDs for each subgroup. Results We included 32 studies (6505 participants); 27/32 (84.38%) in hospitals and none in people's homes. Findings of 14/32 studies (43.75%) were judged, via QUADAS-2, at high overall risk of bias. Fifteen studies measured skin pigmentation and 22 referred only to ethnicity. Compared with standard SaO2 measurement, pulse oximetry probably overestimates oxygen saturation in people with dark skin (pooled mean bias 1.11%; 95% confidence interval 0.29% to 1.93%) and people described as Black/African American (pooled mean bias 1.52%; 0.95% to 2.09%) (moderate- and low-certainty evidence). These results suggest that, for people with dark skin, pulse oximetry may overestimate blood oxygen saturation by around 1% on average compared with SaO2. The bias of pulse oximetry measurements for people with other levels of skin pigmentation, or those from the White/Caucasian group is more uncertain. The data do not suggest overestimation in people from other ethnic groups such as those described as Asian, Hispanic, or mixed ethnicity (pooled mean bias 0.31%, 0.09% to 0.54%), but this evidence is low certainty. Whilst the extent of mean bias is small or negligible for all the subgroups of population evaluated, the associated imprecision is unacceptably large (with the pooled SDs > 1%). Nevertheless, when the extents of measurement bias and precision are considered jointly in Arms, pulse oximetry measurements for all the subgroups appear acceptably accurate (with Arms < 4%). Conclusions Low-certainty evidence suggests that pulse oximetry may overestimate oxygen saturation in people with dark skin and people whose ethnicity is reported as Black/African American, compared with SaO2, although the overestimation may be quite small in hospital settings. The clinical importance of any overestimation will depend on the particular clinical circumstance. Pulse oximetry measurements appear accurate but imprecise for all levels of skin pigmentation. The evidence relates to clinician-measured oximetry in health care environments and may not be reflected in home pulse oximetry where other factors may also influence accuracy.

show abstract

“…Additionally, there is potential for underlying issues regarding the accuracy of medical equipment across different populations. For example, it was shown that pulse oximeters were inaccurate for certain races, which came to light during the COVID crisis, resulting in hidden hypoxemia among patients of colour [18]. This was further investigated, and studies showed that oxygen saturation levels had greater variability in patients who identified as African American, followed by Hispanic, Asian, and lastly, Caucasian patients.…”

Section: Discussionmentioning

confidence: 99%

Vital signs as a source of racial bias

Velichkovska

Gjoreski

Denkovski

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Background: racial bias has been shown to be present in clinical data, affecting patients unfairly based on their race, ethnicity and socio-economic status. This problem has the potential to be significantly exacerbated in the light of Artificial Intelligence-aided clinical decision making. We sought to investigate whether bias can be introduced from sources that are considered neutral with respect to ethnicity and race and consequently routinely used in modelling, specifically vital signs. Methods: to perform our analysis, we extracted vital signs from 49,610 admissions from a cohort of adult patients during the first 24 hours after the admission to the Intensive Care Units (ICU), derived from multi-centre eICU-CRD database and single-centre MIMIC-III database, spanning over 208 hospitals and 335 ICUs. Using heart rate, SaO2, respiratory rate, systolic, diastolic, and mean blood pressure, we develop machine learning models based on Logistic Regression and eXtreme Gradient Boosting and investigate their performance in predicting patients' self-reported race. To balance the dataset between the three ethno-races considered in our study, we use a matching cohort based on age, gender, and admission diagnosis. Findings: standard machine learning models, derived solely on six vital signs can be used to predict patients' self-reported race with AUC of 75%. Our findings hold under diverse patient populations, derived from multiple hospitals and intensive care units. We also show that oxygen saturation is a highly predictive variable, even when measured through methods other than pulse oximetry, namely arterial blood gas analysis, suggesting that addressing bias in routinely collected clinical variables will be challenging. Interpretation: our finding that machine learning models can predict self-reported race using solely vital signs creates a significant risk in clinical decision making, further exacerbating racial inequalities, with highly challenging mitigation measures. Funding: The funders had no role in the design of this study.

show abstract

Analysis of Discrepancies Between Pulse Oximetry and Arterial Oxygen Saturation Measurements by Race and Ethnicity and Association With Organ Dysfunction and Mortality

Cited by 148 publications

References 26 publications

Multispectral snapshot imaging of skin microcirculatory hemoglobin oxygen saturation using artificial neural networks trained on in vivo data

Multispectral snapshot imaging of skin microcirculatory hemoglobin oxygen saturation using artificial neural networks trained on in vivo data

The effects of skin pigmentation on the accuracy of pulse oximetry in measuring oxygen saturation: a systematic review and meta-analysis

Vital signs as a source of racial bias

Contact Info

Product

Resources

About