(769) Impact of Race on Pulse Oximeter Accuracy in Hospitalized Children With Congenital Heart Disease

Introduction: Although the saturation of peripheral oxygen determined by pulse oximetry (SpO2) is assumed to reflect arterial oxygen saturation measured by blood gas analysis (SaO2), studies suggest that the accuracy of SpO2 is lower among Black patients than other racial groups, particularly when SpO2 is relatively low. We studied pulse oximeter accuracy in children with congenital heart disease (CHD) to determine whether the difference between SpO2 and SaO2 is larger among Black children than other races and increases as the SpO2 value decreases.

Methods: This was a retrospective observational study performed via EPIC query using Mount Sinai Data Warehouse. The sample included children < 22 years old with CHD of any type, hospitalized between 2019 and 2021. For each blood gas measurement, a single SpO2 was selected based on the following criteria: patient on same respiratory support, < 10 minutes between the two measurements, and SpO2 value closest to the SaO2 value. For each measurement, pulse oximeter bias was calculated, defined as the difference between SpO2 and SaO2. Per patient, average pulse oximeter bias was used to minimize potential confounders inherent to a single patient. General linear models evaluated the associations among mean pulse oximeter bias, race, and mean SpO2 values, adjusting for the following covariates: sex, age, temperature, hematocrit, pH, and SaO2.

Results: The analysis cohort (N=307) included 58 (18.9%) Black patients. There were a total of 5963 values for pulse oximeter bias. In a model with race (Black versus non-Black children) as the factor of interest, mean pulse oximeter bias was not statistically different between groups when adjusting for covariates (β ̂ = -0.44 [-1.03, 0.16], p=0.15). In a model with mean SpO2 as the factor of interest, there was a statistically significant decrease in the absolute value of mean pulse oximeter bias as mean SpO2 increased when adjusting for covariates (relative change 0.92 [0.90, 0.93], p< 0.0001).

Conclusions: Race was not statistically associated with pulse oximetry accuracy in children with CHD. Absolute value of mean pulse oximeter bias decreased by 8% for every percentage point increase in mean SpO2, confirming the observation that pulse oximetry is less accurate at relatively lower saturation levels.