EquiOx

Year: 2026

Type: article

Abstract: Background Pulse oximeter performance may vary by skin pigmentation, but most data are retrospective with key limitations. Research Question Does pulse oximeter bias in critically ill adults differ by skin pigmentation? Study Design and Methods In this prospective single-center study of 631 patients in the ICU (2022-2024), we directly observed simultaneous pulse oximeter oxygen saturation (Spo 2 ) and arterial blood functional oxygen saturation (Sao 2 ). Skin pigment was assessed using the subjective Monk skin tone scale and objective spectrophotometry measurement (individual typology angle [ITA]). We quantified pulse oximeter bias (mean difference between Spo 2 and Sao 2 ) and average root mean square error (A RMS ) and estimated adjusted effects of skin pigment on bias or A RMS with targeted maximum likelihood estimation. Results Among 1,760 paired measurements, median Sao 2 was 98% (interquartile range [IQR], 96%-99%) with 40 episodes of stable hypoxemia (Spo 2 < 90%). Spo 2 systematically underestimated Sao 2 (median bias, –1.70; IQR, –2.84 to –0.50). Bias was less in patients with darker skin (ITA < –30°; median, –1.05 [IQR, –2.44 to –0.10]) vs lighter skin (ITA > 30°; median, –2.01 [IQR, –3.34 to –1.00]) and remained significantly different in adjusted analyses. A RMS was 3.87 (95% CI, 3.25-4.53) overall and 4.49 (95% CI, 2.63-7.07) in patients with darker skin. Interpretation Our results show that in this large ICU cohort, hypoxemia was rare, US Food and Drug Administration-cleared pulse oximeter Spo 2 systematically underestimated Sao 2 , and performance varied by skin pigment. Bias was less in patients with darker skin. Pulse oximeter inaccuracies in patients in the ICU may be more substantial than clinicians recognize. Additional real-world studies with multiple oximeter brands and more stable hypoxemia are needed to understand bias variation with skin pigment further. Clinical Trial Registration ClinicalTrials.gov; No.: NCT05554510; URL: www.clinicaltrials.gov

Source: CHEST Critical Care

DOI: 10.1016/j.chstcc.2026.100305

Institutions: University of California, San Francisco, Northwestern University, Washington University in St. Louis, Harvard University, NYU Langone Health, New York University

Language: English

Topic: Non-Invasive Vital Sign Monitoring

Subfield: Biomedical Engineering

Field: Engineering

Domain: Physical Sciences

SDGs: Good health and well-being, Reduced inequalities

Open Access status: gold

Estimated APC paid: $2,500

Funders: U.S. Food and Drug Administration, Center of Excellence in Regulatory Science and Innovation, National Institute of Mental Health, Office of Minority Health

Award: G4883553062

OpenAlex ID: W7172269870