Racial disparities in pulse oximetry: measurement bias and clinical implications—a systematic review

Introduction

Pulse oximetry is one of the basic clinical monitoring devices used across the hospital: emergency, intensive care, surgery, and increasingly in home health settings (1). Its value lies in its ability to provide continuous, non-invasive measurements of arterial oxygen saturation (SpO₂). The device operates according to the modified Beer-Lambert law, which links light absorption to the concentration of haemoglobin species within a defined path length (2). Red light at about 660 nm and infrared light at about 940 nm pass through tissue, and changes in absorbance reflect the balance between oxyhaemoglobin and deoxyhaemoglobin (3). Common sensor sites include the fingertip, earlobe, and toe, where a probe containing LEDs and a photodetector measure transmitted or reflected light (4, 5).

The COVID-19 pandemic increased reliance on pulse oximetry for remote triage and home monitoring (6). During this period, concerns about accuracy in individuals with darker skin tones gained new attention, supported by evidence that the device may overestimate oxygen saturation in this population (7, 8). Because SpO₂ readings often determine the need for hospital admission, oxygen therapy, and access to specific treatments, even small inaccuracies can reinforce health inequities (9, 10).

The roots of this problem were noted decades ago, yet systematic investigation was limited until recently. Souza-Barros suggested that melanin absorbs more red light than infrared light, and this can interfere with the device's algorithms, especially at lower saturation levels where decisions carry higher risk (11). Recent reviews confirm that accuracy decreases in individuals with dark skin, with the greatest errors seen at saturations below 90% (12). These inaccuracies can lead to occult hypoxaemia, where true oxygen levels are significantly lower than the SpO₂. Sjoding and colleagues (2020) showed that dark-skinned patients experienced nearly three times the rate of occult hypoxaemia compared to white patients. Additional studies show that these inaccuracies are linked to delayed recognition of deterioration and reduced access to needed interventions (13, 14). Wong and colleagues (2021) also found that errors were more common in Black and Hispanic patients, and these were associated with organ dysfunction and higher in-hospital mortality (15).

The clinical impact spans acute and long-term care. In critical care, SpO₂ informs decisions about mechanical ventilation and escalation of oxygen support or ventilation support (16). In COVID-19, inaccuracy contributed to delayed eligibility for therapies such as remdesivir and dexamethasone among Black and Hispanic patients (9). Outside acute care, pulse oximetry supports decisions about home oxygen therapy, chronic respiratory disease management, and preoperative risk assessment (17). Studies also show that black patients may receive lower oxygen delivery rates in intensive care units for similar presentations because of inaccurate SpO₂ readings (16).

The wider implications of this bias intersect with concerns about structural racism in healthcare. When essential medical devices function less effectively for some populations, they can deepen existing disparities (18, 19). These concerns are especially relevant in conditions where oxygenation plays a central role, such as COVID-19, COPD, and cardiovascular disease (20).

Despite growing awareness, significant gaps remain. Studies differ in how they measure or classify skin tone, the device models they test, and the clinical settings they evaluate (21). Evidence on the accuracy of outpatient and home monitoring is limited. Device-specific comparisons are also incomplete, although early findings suggest that some sensor configurations may be more vulnerable to bias (10). Research across different COVID-19 waves and vaccination phases shows that bias persists in these contexts (14, 22). These issues have prompted regulatory concern. The FDA has initiated advisory committee review, and lawmakers have urged an evaluation of the effects of skin tone on device performance (23, 24).

This review synthesised existing evidence on racial disparities in pulse oximeter accuracy, quantified the degree of bias, evaluated clinical effects, highlighted implications for practice, and identified research priorities needed to achieve equitable device performance. While prior systematic reviews have examined pulse oximeter accuracy across varying skin pigmentation levels, most have focused primarily on aggregate accuracy metrics without integrating downstream clinical consequences or anaesthesiology-specific implications.

In contrast, this review synthesises paired SpO₂–SaO₂ comparisons stratified by race or skin tone and explicitly links measurement bias to occult hypoxaemia, treatment escalation thresholds, and perioperative decision-making. By integrating device-level performance characteristics with clinical risk pathways, this review aims to provide actionable insights for anaesthesiologists and critical care clinicians.

Methodology

This systematic review was reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines to ensure methodological transparency and rigour, as shown in Figure 1. The review protocol is registered in the open science framework (ref ID: 10.17605/OSF.IO/F8894. The review followed the Joanna Briggs Institute (JBI) recommendations for conducting a meta-aggregative review of mixed-method studies. The process involved the following key steps: identification of research questions, development of a search strategy, study selection, data extraction, critical appraisal, data synthesis, interpretation of results, and reporting of findings.

Research questions

The following research questions guided this review.

• How is pulse oximetry accuracy reported across Black and White patients when compared to arterial blood gas as the reference standard?
• What patterns emerge regarding the magnitude and direction of SpO₂–SaO₂ discrepancies across racial or skin-tone groups?
• How is occult hypoxaemia described in studies involving Black patients compared with White patients, and what clinical consequences are reported?
• What technical, physiological, or contextual factors are identified as contributing to disparities in pulse oximeter accuracy?
• In what ways do individual patient characteristics (e.g., comorbidities, haemoglobin levels, skin tone intensity) interact with device performance to influence measurement accuracy?

Search strategy

A comprehensive and structured search was conducted to retrieve all potentially relevant studies on pulse oximeter accuracy across racial groups, with particular focus on Black and White populations. The search aimed to identify peer-reviewed and grey literature reporting on diagnostic accuracy, measurement bias, or clinical implications related to pulse oximetry stratified by race or skin pigmentation.

The search combined controlled vocabulary (e.g., MeSH terms) and free-text keywords under five conceptual domains with the combination #1 AND #2 AND #3:

Searches were conducted in the following electronic databases: PubMed, Scopus, Web of Science, CINAHL Complete, and the WHO Global Index Medicus. Each database was accessed through its respective interface: PubMed via NCBI; Scopus via Elsevier; Web of Science via Clarivate; CINAHL Complete via EBSCOhost; and WHO Global Index Medicus via https://www.globalindexmedicus.net.

Grey literature was sourced from the WHO Global Index Medicus, and manual searches of reference lists of included articles were also performed to identify additional studies. Consensus meetings were held among the research team to resolve any disagreements regarding the search. The initial search was conducted in June 2024 and was repeated after twelve months to ensure currency.

Study selection

Eligibility criteria guided the screening and inclusion of studies.

Inclusion criteria

Studies were included if they met the following criteria:

• Population: Human studies of any age group reporting original data, including subgroup analyses by race, ethnicity, or skin pigmentation.
• Comparator/Reference Standard: Studies comparing pulse oximetry with arterial blood gas (ABG) as the reference standard.
• Outcomes: Studies reporting SpO₂–SaO₂ discrepancies or occult hypoxaemia stratified by race or skin tone (e.g., Fitzpatrick scale).
• Study Design: Randomised controlled trials and observational diagnostic accuracy studies (cohort, cross-sectional, quasi-experimental).
• Time Frame: Studies published up to June 2025.
• Language: English-language publications.

Exclusion criteria

Studies were excluded if they were:

• Animal or in vitro studies
• Without subgroup analysis by race or skin tone
• Reviews, meta-analyses, editorials, or commentaries without primary data
• Non-English studies without full English translations.

Study selection and screening

Study screening was conducted in Rayyan in three sequential stages. First, duplicate records were identified and removed. Second, two reviewers independently screened titles and abstracts against the predefined eligibility criteria. Third, full texts of potentially eligible records were assessed to confirm inclusion. Any disagreements were resolved by consensus through discussion and, where needed, adjudicated by a third reviewer.

To reduce selection bias, screening was performed in a blinded format in Rayyan, with author names, institutional affiliations, and journal identifiers concealed so that inclusion decisions were based solely on content relevance.

Search strategy and terms

We used a combination of controlled vocabulary (e.g., MeSH where applicable) and free-text terms, structured around four concept blocks: (1) pulse oximetry/oxygen saturation [“Oximetry”(MeSH) OR pulse oximeter* OR oxygen saturation OR SpO₂ OR oximeter accuracy], (2) race/ethnicity/skin pigmentation [“Ethnicity”(MeSH) OR race OR racial bias OR skin pigmentation OR skin tone OR Black OR White OR African descent OR Caucasian], (3) measurement accuracy and bias [“Data Accuracy”(MeSH) OR measurement bias OR diagnostic accuracy OR overestimation OR underestimation OR error rate OR validation study], and (4) No study design filter was applied at the search stage to maximise sensitivity. Study design eligibility criteria were applied during screening.

We excluded non-human studies and non-target paediatric populations using the following exclusion terms: NOT (animal study OR veterinary). Studies conducted exclusively under extracorporeal circulation (e.g., ECMO) were excluded if extracorporeal flow fundamentally altered pulsatile signal acquisition.

Data extraction

Data were extracted using a structured extraction form developed a priori and piloted on three included studies to ensure clarity and consistency. Extracted variables included study identification and setting (author, year, country), design and population characteristics (study design, sample size, age, sex/gender, ethnicity/race categories), index and reference measures (SpO₂ and SaO₂, including reported means and/or paired comparisons), device characteristics (pulse oximeter brand/model where reported), analytic approach (adjustments and potential confounders), and outcomes (magnitude and direction of SpO₂–SaO₂ discrepancy, prevalence of occult hypoxemia, and clinically relevant downstream effects where reported).

Two reviewers independently extracted data. Discrepancies were resolved by consensus, with adjudication by a third reviewer when necessary. Where feasible, extraction and verification were performed without reference to study identifiers to minimise potential bias.

Critical appraisal

Methodological quality was assessed using the Joanna Briggs Institute (JBI) critical appraisal tools, selecting the checklist appropriate to each study design. Appraisal focused on domains relevant to internal validity, including participant selection, measurement of exposure/outcomes, management of confounding, completeness of data, and appropriateness of statistical analyses. Quality assessments informed the interpretation of the findings and the certainty of evidence.

Grading the certainty of evidence

Certainty of evidence was evaluated using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework. Certainty ratings were assigned at the outcome level across the domains of risk of bias, inconsistency, indirectness, imprecision, and publication bias. Because no quantitative meta-analysis was performed, domain assessments were conducted narratively based on study design, heterogeneity, and reporting characteristics.

Data synthesis and reporting

Given heterogeneity in study designs, clinical settings, definitions of occult hypoxemia, and reporting of accuracy metrics, results were synthesised primarily through narrative synthesis, with a structured presentation of quantitative findings. Heterogeneity was substantial due to differences in definitions of occult hypoxaemia, device models, patient populations, perfusion states, measurement timing, and inconsistent reporting of effect measures. Given this variability, statistical pooling was not performed to avoid potentially misleading summary estimates.

Where reported, effect estimates describing SpO₂–SaO₂ discrepancy and occult hypoxemia prevalence were summarised and compared across racial/ethnic groups and clinical contexts. In parallel, a thematic synthesis was undertaken to identify recurring patterns, including the direction and magnitude of measurement bias by race/skin pigmentation, the frequency of occult hypoxemia, and contextual factors (e.g., device calibration, perfusion, clinical acuity) that may influence accuracy. Reporting followed PRISMA principles to provide a transparent account of study selection, appraisal, and synthesis.

Results

Fourteen studies met the inclusion criteria and were synthesised narratively (Tables 1–3). The evidence base was heterogeneous in design, setting, and outcome definitions, which limited direct cross-study comparability. Most included studies were observational (retrospective cohort or cross-sectional), with a smaller number of prospective validation/comparative studies and one commentary/review included for contextual synthesis.

Table 1. Study characteristics and measurement bias findings.

This table presents a systematic overview of fifteen key studies examining pulse oximetry accuracy across different racial and ethnic groups, revealing critical disparities in device performance.

Table 2. Patterns of bias and influencing factors.

This table provides essential methodological context that helps interpret the validity and applicability of the findings.

Table 3. Risk of bias and appraisal of included studies.

This table provides a critical appraisal of the methodological quality of included studies, which is essential for determining the strength of evidence.

.

Study characteristics

Studies spanned adult and paediatric populations, including neonatal/infant cohorts, general hospital inpatients, ICU cohorts, and COVID-19–specific cohorts. Sample sizes ranged from small, controlled volunteer studies (n = 20) to large hospital datasets involving tens of thousands of patients (e.g., n = 40,738; n = 10,698). Race/ethnicity was mostly captured using self-identified categories (typically Black and White, with some studies including Hispanic, Asian, and other groups). A minority explicitly examined skin tone using scales (e.g., Fitzpatrick) or broad descriptors such as “dark” vs. “light” skin (Table 2). Across studies, pulse oximeter saturation (SpO₂) was compared against arterial oxygen saturation (SaO₂) obtained via arterial blood gas/co-oximetry as the reference standard. Accuracy was reported using mean bias, error distributions (±SD), or RMS error, and several studies assessed occult hypoxaemia. Eleven studies stratified findings by self-identified race or ethnicity, whereas three studies utilised objective pigmentation scales (e.g., Fitzpatrick classification). Because race is a social construct and melanin concentration influences optical absorption properties, findings were interpreted cautiously across these measurement approaches.

Measurement bias and accuracy findings

Across the included evidence, a recurring pattern was observed: pulse oximetry tended to overestimate arterial oxygen saturation in darker-skinned and/or Black patients, particularly in hypoxaemic ranges. Several large retrospective cohorts reported positive bias (SpO₂ higher than SaO₂) among Black patients relative to White patients, with mean differences typically small in absolute terms (often around 1–2.5 percentage points). In COVID-19 cohorts, overestimation was also reported in Black and Hispanic groups compared with non-Hispanic White patients, suggesting that even modest bias may influence threshold-based decisions (e.g., oxygen escalation or eligibility for time-sensitive therapies).

Some investigations reported low overall bias, minimal differences by race/skin tone, or findings interpreted as not clinically significant, particularly in cohorts with higher average saturations or where measurement protocols were more controlled. Nevertheless, the presence of null or mixed results appeared to reflect context dependence rather than the absence of bias, as several studies suggested performance varies by oxygenation level, device characteristics, and patient physiology.

Where stratified analyses were available, measurement error increased as true SaO₂ decreased, indicating reduced accuracy in clinically important hypoxaemic ranges. This is particularly relevant because the consequences of misclassification are greatest when true saturation is low and clinical decisions are most sensitive to thresholds. In paediatric cardiac ICU cohorts, discrepancy worsened with hypoxaemia, and Black children demonstrated higher overestimation than White children, supporting the interpretation that bias may be amplified in high-risk physiologic states.

Occult hypoxaemia

Occult hypoxaemia was assessed in multiple studies, most defined as SaO₂ below a clinical threshold (often <88%) despite SpO₂ readings at or above that threshold. Although definitions varied, the direction of findings was consistent: occult hypoxaemia occurred more frequently among Black patients than White patients in several cohorts, including large multicentre analyses. In neonatal cohorts, occult hypoxaemia was also more common in Black infants than White infants, though statistical significance was not consistently demonstrated. Device-comparison and pathway-focused studies indicated that occult hypoxaemia can occur at a meaningful frequency and may be influenced by device selection and calibration differences.

Definition variability, and timing considerations in occult hypoxaemia

Across included studies, occult hypoxaemia was variably defined, most commonly as SaO₂ < 88% despite SpO₂ ≥ 88%, although some studies used thresholds of <90% or context-specific cutoffs. These definitional differences influence reported prevalence estimates and limit direct cross-study comparability.

Additionally, the temporal relationship between pulse oximetry measurement and arterial blood sampling varied. In some studies, measurements were taken simultaneously, while in others, short but unspecified time intervals were allowed. Even brief delays may introduce misclassification in unstable or rapidly deteriorating patients. This heterogeneity supports the decision to employ narrative rather than pooled meta-analytic synthesis and highlights the need for future prospective studies using standardised definitions and synchronised measurements.

Factors influencing bias and heterogeneity

Table 2 highlights recurring contributors to variability in findings: skin pigmentation-related optical interference, manufacturer calibration and device variability, probe placement and motion artefact, and physiologic confounders such as low perfusion states or vasoconstriction. Disease contexts (e.g., COVID-19 pneumonitis, congenital heart disease, ICU shock) likely exacerbated performance limitations, potentially increasing the clinical impact of bias during hypoxaemia. In a randomised study of 20,802 surgical patients, pulse oximetry increased the detection of hypoxemia 19-fold in the operating room and PACU (P < 0.00001) and prompted significant changes in postoperative management, particularly in patients with higher ASA physical status. These findings highlight that accurate pulse oximetry is critical in perioperative patients, as delayed or inaccurate detection of hypoxemia may postpone timely intervention in the PACU.

Risk of bias

Most studies were appraised as low-to-moderate risk of bias (Table 3). Key limitations included retrospective design, inconsistent reporting of device models/sensor sites, timing differences between SpO₂ and arterial sampling, and non-uniform definitions of occult hypoxaemia. Despite these constraints, the overall synthesis indicates measurable racial/skin-related differences in pulse oximeter performance, with the most clinically significant implications occurring in hypoxaemic patients and threshold-driven care pathways.

Discussion

This review synthesised evidence from multiple studies examining racial and skin tone differences in pulse oximetry accuracy, with a particular focus on discrepancies between arterial oxygen saturation (SaO₂) and pulse oximeter readings (SpO₂). The findings revealed consistent but variable racial disparities across studies, particularly between black and white patients. Overall, pulse oximeters tend to overestimate oxygen saturation levels in individuals with darker skin pigmentation, especially in low perfusion or hypoxaemic conditions. Although the magnitude of this bias was generally small, its clinical implications were significant because even modest overestimations could delay recognition and treatment of hypoxaemia.

The studies also identified a complex interplay among physiological, technical, and contextual factors, such as device calibration, perfusion index, comorbidities, and environmental conditions, that contributed to measurement variability.

Pulse oximeters use red (about 660 nm) and infrared (about 940 nm) light and apply the modified Beer–Lambert law to estimate oxygen saturation (26). Melanin also absorbs light in these wavelength ranges, so darker skin can change the light ratios that the device algorithm uses, and most commercial oximeters were originally calibrated mainly in light-skinned populations (27, 28). Accuracy is usually reported as a root mean square (RMS) error, which can hide a consistent overestimation of SpO₂ in darker-skinned patients, especially at lower SaO₂ levels (27, 28). Additional factors—such as low perfusion, shock, vasoconstriction, motion, probe placement, and sensor design—can further magnify these effects, so the observed racial differences in accuracy reflect the interaction of optics, calibration data, and physiology rather than race as a biological trait (7, 27).

Individual characteristics, including skin tone, circulatory status, and underlying health conditions, were shown to influence device performance. While some studies demonstrated minimal or non-significant differences, others revealed statistically meaningful discrepancies, particularly in critical care and paediatric settings. The evidence collectively underscores the need for inclusive calibration and standardised validation protocols across racially diverse populations.

Evidence consistently indicates that pulse oximeters demonstrate reduced accuracy in patients with darker skin tones, particularly among black populations, compared to their white counterparts. Sjoding et al. reported that pulse oximetry overestimated arterial oxygen saturation in black patients, leading to missed detection of hypoxaemia (7). Similarly, Vesoulis et al. found that this estimation extended to younger populations, with black infants experiencing greater discrepancies in SpO2-SaO2 measurements (19). These findings highlight a systemic limitation in device calibration, as most oximeters were historically validated on lighter-skinned individuals, leading to biased accuracy when used across diverse populations.

Malar et al. corroborated these results by demonstrating small but measurable race-based differences in SpO2-SaO2 accuracy, showing that Asian and black patients exhibited slightly greater mean errors than white patients (29). Sharma et al. further highlighted that accuracy varied with perfusion index (PI), with greater deviations occurring under low perfusion conditions, suggesting that physiological variables can exacerbate the effects of pigmentation on sensor performance (30). Coghill et al. extended these findings to paediatric cardiac patients, noting that pulse oximeters consistently overestimated oxygen saturation in black children, particularly during hypoxaemia episodes (31).

Despite variability across studies, the consensus points to a persistent racial discrepancy in pulse oximeter accuracy. Even when differences were minor, as in Wiles et al., they remain clinically relevant given that oxygen therapy decisions often hinge on narrow saturation thresholds (10). Systematic reviews and experimental work have demonstrated pigmentation-related differences in oximeter performance, while Okunlola et al. highlighted the continued inconsistency of device accuracy across skin tones (11, 12, 26–28, 32). Collectively, these findings confirm that while pulse oximeters are valuable clinical tools, their accuracy is not uniform across racial groups, reinforcing the need for more inclusive technological validation.

The magnitude and direction of SpO2-SaO2 discrepancies varied across studies but consistently indicated a positive bias in darker-skinned populations. Sjoding et al. found that pulse oximeters overestimated SaO₂ in black patients with discrepancies large enough to influence clinical outcomes (7). Malar et al. quantified this bias, showing that black and Asian participants had greater mean errors than white participants, though the magnitude of error was modest (29). Wiles et al. observed similar findings in COVID-19 patients reporting small but consistent overestimations among non-white groups (10).

These discrepancies, though numerically small, have meaningful clinical consequences. For instance, even a 2% overestimation can mask hypoxaemia that would otherwise necessitate medical intervention. Sharma et al. observed that bias increased under low perfusion conditions, suggesting that the interaction between device performance and physiological state can amplify measurement errors (30). Coghill et al. confirmed these effects in paediatric critical care, where black children displayed greater overestimation of oxygen levels during episodes of true hypoxaemia (31).

Recent evidence also suggests that disparities may extend beyond measurement error alone. Matos et al. demonstrated racial and sex differences in the utilisation of arterial blood gas testing across U.S. hospitals (32). When pulse oximeter overestimation occurs in populations already less likely to receive confirmatory ABG testing, the combined effect may compound delayed recognition of hypoxaemia (32).

Okunlola et al. reinforced that dark pigmentation contributes to overestimation patterns across devices, though the extent of bias differed among brands and models (33). The consensus across studies suggests that the direction of the discrepancy is consistently upward, with pulse oximeters tending to over-read oxygen saturation in darker-skinned individuals, highlighting a systemic calibration limitation in current device designs.

Clinical consequences of occult hypoxaemia are among the most important findings of this review. Sjoding et al. found that black patients who experienced occult hypoxaemia (SaO₂ below 88% despite normal SpO₂ readings) were significantly more than white patients (7). This phenomenon can lead to delays in oxygen therapy and worsen clinical outcomes. Vesoulis et al. similarly reported higher rates of occult hypoxaemia in Black preterm infants, emphasising that these disparities extend beyond adults and into neonatal populations (25).

Malar et al. observed a lower overall incidence of occult hypoxaemia but confirmed that small overestimations in oxygen saturation could translate to under recognition of hypoxia in clinical practice (29). Fawzy et al. provided critical evidence linking this measurement bias to real-world outcomes in COVID-19 patients, in which overestimated oxygen saturation led to delayed treatment initiation and potentially higher morbidity (34). Coghill et al. noted comparable risks in paediatric cardiac intensive care settings, where overestimation in black patients masked episodes of desaturation (31).

These findings collectively highlight the clinical risk of relying solely on pulse oximetry for oxygen assessment, especially in racially diverse populations. The evidence underscores the importance of confirmatory testing, such as arterial blood gas analysis, in situations where accurate assessment of oxygen is essential. Okunlola et al. reinforced that unrecognised hypoxaemia can exacerbate morbidity and mortality, particularly in patients with darker skin, making equitable device accuracy a pressing public health issue (33).

Beyond racial differences, studies identified multiple technical, physiological, and contextual factors that exacerbate or mediate measurement inaccuracies. Marlar et al. showed that pulse oximeter performance varies by race and by measured skin pigmentation, suggesting that current device calibration and sensor characteristics are not optimised equally across all skin tones (29). Other studies and regulatory guidance highlight that low perfusion, motion artefacts, and probe placement can reduce pulse oximeter accuracy, particularly in patients with darker skin (2, 4, 5, 23, 30).

Physiological factors, such as anaemia, hypothermia and vasoconstriction, were also identified as confounders. Okunlola et al. demonstrated that calibration limitations in certain devices contributed to higher measurement errors under these conditions (33). Stell et al. showed that bias levels differed across pulse oximeter brands, underscoring the importance of device-specific performance evaluations (35). The Singapore study reported that jaundice and bilirubin levels can alter optical absorption, adding another layer of variability in patients with darker pigmentation.

Fawzy et al. highlighted contextual influences such as comorbidities, disease, severity, and demographic differences, which interact with the device performance (34). Together, these studies reveal that measurement disparities arise from a combination of technological, biological, and situational factors. Consequently, achieving true equity in pulse oximetry requires not only hardware adjustments but also broader clinical awareness of the multifactorial influences at play.

Individual-level factors, including pigmentation, perfusion, and physiological state, significantly influence pulse oximetry readings. Experimental work and systematic reviews show that higher skin pigmentation alters light absorption and can contribute to overestimation of oxygen saturation in individuals with darker skin (11, 12, 26–28, 33). Sharma et al. further demonstrated that perfusion index strongly modulates measurement accuracy, especially in racially diverse patients and at low perfusion levels (30).

Fawzy et al. highlighted how age, comorbidities, and disease severity can further affect reliability, with older adults and critically ill patients exhibiting more variable readings (34). In paediatric contexts, Coghill et al. identified developmental differences and unique circulatory patterns as contributors to reading variability among black and white children (31). Malar et al. concluded that race-related differences in oxygen saturation accuracy arise from the combined effects of pigmentation, physiology, and calibration limitations rather than race alone (29).

These findings emphasise the need for individualised interpretation of pulse oximetry data. Clinicians should consider patient-specific variables when interpreting SpO₂ readings, particularly for individuals with darker skin tones or compromised perfusion.

Recommendations for clinical practice

Healthcare providers must exercise heightened vigilance when interpreting pulse oximetry readings in patients with darker skin pigmentation. Clinical decision-making should not rely on pulse oximetry alone but should incorporate comprehensive assessment, including clinical examination, patient symptoms, and respiratory rate. Lower thresholds for obtaining confirmatory arterial blood gas measurements are warranted, particularly when pulse oximetry readings are borderline or inconsistent with clinical presentation. Hospital systems should develop protocols that acknowledge device limitations and ensure equitable access to supplemental oxygen and advanced respiratory support.

Synthesis across all tables

Collectively, these tables show a consistent pattern of pulse oximetry bias across racial and ethnic groups, with potentially serious clinical implications. Pulse oximeters frequently overestimate oxygen saturation in patients with darker skin pigmentation, resulting in occult or hidden hypoxaemia in approximately 7%–24% of cases, depending on the study and population. This systematic presentation of evidence provides a foundation for understanding a critical health equity issue in clinical practice and underscores the urgent need to improve pulse oximeter accuracy across all patient populations.

Study design strengths, limitations, and need for prospective research

Most included studies are retrospective observational analyses. This predominance provides real-world evidence with large sample sizes that reflect actual clinical practice, but it also introduces limitations, including incomplete data, unmeasured confounders, and selection bias.

The absence of randomised controlled trials is notable and reflects the ethical challenges of experimentally manipulating oxygen levels to study device accuracy. These findings underline the need for prospective studies with standardised protocols, including predetermined thresholds for hypoxaemia, consistent device selection, objective measurement of skin pigmentation, and careful control of confounding variables. Future research should prioritise objective pigmentation metrics alongside race to disentangle biologic optical effects from socioclinical constructs.

Exclusion of ECMO-based cohorts may limit generalisability to extracorporeal support populations, where perfusion dynamics differ substantially.

Research gaps and recommendations for future studies

The tables reveal important research gaps: limited representation of some ethnic groups, inconsistent methodologies, variable device types, and insufficient prospective validation studies. Future research on racial disparities in pulse oximetry measurements should standardise definitions and methodologies to enable consistent comparison across studies, particularly in low- and middle-income countries where data are limited. Longitudinal studies with extended follow-up periods are needed to understand variation in pulse oximeter measurements over time and in poorly perfused states. A variety of study designs—including qualitative and cohort studies—should be used to capture the outcomes and experiences associated with inaccurate pulse oximeter readings. Collaborative global efforts are essential to develop well-calibrated devices suitable for individuals from different racial and ethnic backgrounds.

Key clinical implications and recommendations for practice

In perioperative and critical care anaesthesiology, pulse oximetry directly informs decisions on oxygen titration, ventilatory support, and escalation. During intraoperative hypoperfusion or vasoconstrictive states, signal accuracy may be compromised, and falsely reassuring readings in the PACU—where supplemental oxygen is routinely administered—may delay recognition of clinically significant hypoxemia in high-risk surgical patients.

The risk assessment and evidence synthesis have clear implications for clinical practice:

• Clinicians should maintain heightened awareness of potential pulse oximeter inaccuracy in patients with darker skin pigmentation.
• Clinical decision-making should not rely on pulse oximetry alone but should incorporate comprehensive assessment, including clinical examination, patient symptoms, and respiratory rate.
• Lower thresholds for obtaining confirmatory arterial blood gas measurements may be warranted, especially when pulse oximetry readings are borderline or inconsistent with the clinical picture.
• Hospital systems should develop protocols that explicitly acknowledge device limitations and ensure equitable access to supplemental oxygen and advanced respiratory support.
• Device manufacturers may need to recalibrate algorithms using diverse populations to improve accuracy across racial and ethnic groups.

Comparison with prior systematic reviews and novel contribution

Previous systematic reviews have evaluated pulse oximetry accuracy across skin pigmentation groups, often including broad validation studies or device-level assessments without clinical outcome analysis (12, 26–28, 32). In contrast, this review focused specifically on paired SpO₂–SaO₂ comparisons with racial stratification and explicit reporting of occult hypoxaemia. By integrating measurement bias with downstream clinical decision-making pathways—particularly in anaesthesiology and critical care contexts—this review extends prior work and provides a more clinically actionable synthesis of the evidence.

Conclusion

Evidence consistently demonstrates small but clinically meaningful racial disparities in pulse oximetry accuracy, with a greater risk of occult hypoxaemia among patients with darker skin pigmentation. While absolute bias is modest, threshold-based clinical decision-making may amplify its impact in critical care and anaesthesiology contexts. Heterogeneity across study designs and definitions limits certainty, underscoring the need for standardised methodology, inclusive device calibration, and rigorous validation across diverse populations. Addressing measurement bias is essential for equitable patient safety and more just clinical decision-making.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. Generative AI disclosure: The authors used OpenAI ChatGPT (GPT-5.2 Thinking; OpenAI) to assist with language editing and formatting of the manuscript. All outputs were reviewed, verified, and revised by the authors, who remain fully responsible for the content.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Lalit Gupta, University of Delhi, India

Reviewed by: Devang Bharti, University of Delhi, India

Guo Tony, National Taiwan University, Taiwan