Pulse Oximetry Error in Patients with Single Ventricle Palliation Increases as a Function of Systemic Hypoxemia and Patient Race

Background: Arterial oxygen saturation monitoring is fundamental to the management of children with cyanotic congenital heart disease (CCHD), particularly those with single ventricle physiology. Pulse oximetry (SpO2) is universally used as a non-invasive estimate of arterial oxygen saturation (SaO2). Darker skin color and lower SaO2 have previously been reported to negatively affect the accuracy of SpO2, but the relationship between self-identified race and SpO2 accuracy has not been investigated in children with CCHD. This study describes the effect of race on SpO2 accuracy over the entire range of arterial saturations experienced by children with CCHD.

Methods: We conducted a single center, retrospective review of children with CCHD undergoing cardiac catheterization under general anesthesia any time prior to a Fontan palliation or biventricular repair. Time-matched SpO2-SaO2 pairs were collected from the procedure log and race identifiers were collected from the STS and IMPACT clinical databases. Self-identified African Americans (AA) were compared to those without AA race. Linear mixed effect modeling was used to describe the relationship between AA race, SaO2, and SpO2. The model included a random effect for patients and an interaction term to determine if AA race affects the relationship between SaO2 and SpO2. In the final models, SaO2 was converted to percent below 100% to facilitate interpretation. The intercept in this context is equal to the expected SpO2 when the SaO2 is 100%.

Results: 123 patients comprising 163 unique catheterizations and 425 SpO2-SaO2 pairs were included in the analysis. 29 subjects (24%) identified as AA. 136/163 catheterizations (83%) occurred during their stage 1 palliation. The median difference between SpO2 and SaO2 was +1%, with interquartile range of -1% to +4%. The SpO2-SaO2 difference increased at lower SaO2 (B=-0.39, p<0.001). There was significant interaction between AA race and SaO2 in the model of SpO2 (p < 0.001). This interaction persisted even after controlling for other available covariates (p < 0.001). In the final model, comparing AA to non-AA subjects, the intercept was lower (91% vs 96%, p<0.001) and the slope was less negative (-0.46 vs -0.74, p<0.001) demonstrating that in AA patients, SpO2 increasingly overestimates SaO2 at lower SaO2. For example, at an SaO2 of 80% the expected SpO2 for AA subjects was 82% versus 81% for non-AA subjects, while at an SaO2 of 60% the expected SpO2 for AA subjects was 73% versus 66% for non-AA subjects.

Conclusion: These data demonstrate that pulse oximetry, a mainstay for clinical decision making in children with cyanotic congenital heart disease, significantly underestimates the degree of critical hypoxemia in African American patients. This failure to accurately detect critical hypoxemia in AA patients with CCHD may contribute to disparities in outcome.