Abstract Rationale Pulse oximetry (SpO2) is widely used to assess oxygenation in acute and critical illness, yet emerging evidence suggests that device performance varies by skin pigmentation and race/ethnicity. Overestimation of oxygen saturation in racial and ethnic minority patients may delay recognition of hypoxemia and contribute to inequities in respiratory care. The Blood-gas and Oximetry Linked Dataset (BOLD), which contains 49,000 paired arterial blood gas (SaO2) and simultaneous SpO2 values across multiple large critical care databases, provides a unique opportunity to quantify physiologic and demographic contributors to hypoxemia. Objectives To evaluate racial and ethnic disparities in the accuracy of pulse oximetry by comparing SpO2-SaO2 discrepancies across groups and their in-hospital outcomes. Methods We conducted a retrospective analysis of adult ICU encounters in BOLD. We included all paired SpO2-SaO2 measurements with temporally aligned arterial blood gases (ABGs). Race and ethnicity were categorized as White, African American/Black, Asian/Pacific Islanders, Hispanic/Latino, and American Natives. The primary outcome was SpO2-SaO2 bias (SpO2 minus SaO2) and its correlation with arterial partial pressure of oxygen (PaO2). Spearman’s correlation and Scatter plots were used for statistical analysis. Results The study population included 49,093 adult individuals admitted in ICU with a mean age of 64.43 (15.80) years. About 56.3% were males, 76.1% White, 9.7% Black, 4.3% Hispanic, 1.8% Asians, and 0.8% were native Americans. In-hospital mortality reported was 17.4% with a mean hospital length of stay 11.55 (13.66) days and mean ICU stay 5.23 (7.13) days. Among the study cohort, African American/Black race showed significantly greater positive SpO2-SaO2 bias compared with White patients and other racial groups. Black patients demonstrated the largest mean difference overestimation of SaO2: +1.76 (4.43) followed by Hispanic patients: +1.49 (3.74). ICU length of stay was higher in Asian and Black patients (6 days, p 0.001), as well as total hospital length of stay was higher in Black and Hispanic population (13.3 days, p 0.001). No difference in hospital mortality was observed among gender (p = 0.117) or racial differences (p = 0.775). Conclusions In a large multicenter ICU dataset with precisely paired SpO2 and SaO2 measurements, we identified systematic racial and ethnic disparities in pulse oximetry accuracy. Minority patients experienced substantially higher variability of differences between arterial and pulse saturations, independent of physiologic modifiers. These findings highlight the need for improved device calibration, cautious clinical interpretation of SpO2 in high-risk groups, and further investigation into equity-focused approaches to oxygenation assessment in critical illness. This abstract is funded by: N/A
Yousaf et al. (Fri,) studied this question.