Performance Inconsistency of Pulse Oximeters Over Five Years of Clinical Use: A Single-Center Study of Masimo Devices

Abstract Rationale Clinical recognition and treatment of hypoxemia rely on pulse oximetry. Despite general awareness of pulse oximeter shortcomings, clinicians may assume consistent performance across time, models, and batches, an assumption essential for applying clinical trial evidence and guidelines at the bedside. Objectives Assess whether the difference between the estimated arterial oxygen saturation from a medical-grade Masimo pulse oximeter versus gold-standard concurrent arterial blood gas changed over 5 years of clinical use in intensive care unit (ICU) patients. Methods Patients with ≥1 concurrent arterial oxygen saturation (SaO2) and pulse oximetry (SpO2) in an ICU with high-frequency vital sign data (SpO2 recorded at 2-second intervals) from 2020-2024 were included. Occult hypoxemia (SaO2<88% with SpO2≥92%), and false hypoxemia (SpO2<88% with SaO2≥92%) were examined by calendar year, as were mean oximeter error (SpO2-SaO2), Accuracy Root Mean Square (Arms, the U.S. Food and Drug Administration (FDA)’s preferred measure), and errors >4% and <-4%. Results were stratified by race/ethnicity as recorded in the electronic medical record. An adjusted linear mixed effects model evaluated calendar-year changes. Measurements and Main Results 223,145 ABGs in 15,129 patients were analyzed. Occult hypoxemia and overestimation errors >4% decreased while false hypoxemia and underestimation errors <-4% increased several-fold between 2020-2024. Mean oximeter error shifted from overestimation to underestimation by mid-2022. Arms fluctuated and was consistently above 3-3.5%, the FDA’s threshold for controlled laboratory testing among healthy volunteers, peaking at end of study. Fluctuation was greatest for SaO2<92% and among non-White patients. By 2024, oximeters underestimated oxygen saturation on average in adjusted models for both Black and White patients with SaO2≥92% and showed relative underestimation among Black versus White patients across SaO2 categories. Conclusions The accuracy of the tested medical-grade pulse oximeter changed year-to-year at a large hospital compared to concurrent arterial blood gas saturation, potentially complicating safe clinical decision-making and application of evidence-based medicine. Errors of false hypoxemia became more common, threatening different clinical consequences and necessitating different adaptations than errors leading to occult hypoxemia.