Pulse oximeters were associated with inconsistent errors between 2020 and 2024, independent of blood oxygen saturation and patient demographics, according to results of a study published in the Annals of the American Thoracic Society.

Researchers sought to determine whether the accuracy and precision of pulse oximeters were changing over time using arterial oxygen saturation (SaO2) paired to concurrent peripheral oxygen saturation (SpO2) from Masimo oximeters connected to high-frequency vital sign data of patients in intensive care units (ICU).

Pulse oximetry was assessed with the disposable Masimo signal extraction technology (SET) low noise cabled sensors (LNCS) wrap or LNCS reusable ear clip sensor (TC-I) connected to a GE Carescape B850 monitor. The study included patients admitted to an ICU in San Francisco between January 2020 and September 2024. The researchers recorded data from SpO2 at 2-second intervals.

Our results highlight shortcomings of the current regulatory paradigm, which certifies pulse oximeters based on laboratory testing in healthy volunteers at a single timepoint.

The participants were classified as hypoxemic (SaO2 <88%) or normoxemic (SaO2 ≥92%) leading to occult (or hidden) hypoxemia (SaO2 <88% when SpO2 ≥92%) as the false negative result and false hypoxemia (SpO2 <88% when SaO2 ≥92%) as the false positive. Values between 88% and 92% were indeterminate. The proportion of SpO2-SaO2 pairs with errors (>4% or <-4%) was calculated and the mean and median pulse oximeter error and Accuracy Root Mean Square (Arms) were assessed by calendar year/quarter, stratified by SaO2 and race/ethnicity.

The researchers analyzed 223,145 arterial blood gas measurements in 15,129 patients (15,916 encounters). Mean (SD) ages ranged from 59.6 (15.8) years in participants enrolled in 2020 to 61.6 (15.3) years in those enrolled in 2024. The proportion of women included in the study ranged from 42.2% to 45.0%.

The incidence of occult hypoxemia decreased from 2.1% to 0.6% between 2020 and 2024, while false hypoxemia increased from 1.2% to 3.0%. Similar patterns were observed across all race/ethnicity subgroups. In Black patients, occult hypoxemia decreased from 3.2% to 0.9%, and false hypoxemia increased from 1.4% to 4.3%. A steady decrease occurred in pulse oximeter overestimation errors above 4%, with an increase in pulse oximeter underestimation errors (<-4%; up 2- to 3-fold for non-White patients).

From 2020 to 2024, the mean pulse oximeter error (SpO2-SaO2) varied, with the unadjusted overall mean error changing from an overestimation of 0.77% to underestimation of oxygen saturation in late 2022, reaching a nadir of -0.65% in 2024. Increased fluctuations in error occurred in lower SaO2 categories.

The researchers noted that Arms fluctuated and then peaked in 2024 for non-White patients, reaching 5.0% for Asians, 6.5% for Blacks, and 5.3% for Hispanics. They also found that Arms was consistently greater than the US Food and Drug Administration’s 3% threshold for finger oximeters and 3.5% threshold for ear oximeters. In White patients, Arms was stable (3.8-4.0%) but it worsened in Black patients from 4.7% to 5.3% in 2020 to 2023 and to 6.5% in 2024, with similar results in Asian and Hispanic patients.

The mean error changed significantly from 2020 to 2024 and was most inconsistent at SaO2 under 88% in covariate-adjusted models. Pulse oximeters underestimated oxygen saturation on average for SaO2 at or above 92% in all patients by 2024, and the decrease in mean error was greatest in Black patients.

The study was limited by its single-center design and utilization of a single device from an individual manufacturer. In addition, patients of Asian and Hispanic descent were underrepresented and the researchers could not account for the possibility of individual practice changes by clinicians and nurses.

“Our results highlight shortcomings of the current regulatory paradigm, which certifies pulse oximeters based on laboratory testing in healthy volunteers at a single timepoint,” the study authors stated. “These findings suggest that regulatory oversight should extend to include both testing in real-world patient populations and periodic recertification to ensure sustained device performance over time.”