Hospital Pulse Oximeters Shift from Overestimation to Underestimation

A critical change in diagnostic accuracy for blood oxygen levels poses new challenges for patient safety and the reliability of health data, particularly for individuals with diverse skin tones.

By Sabin · Wellness & AI3 min read
AI News
Hospital Pulse Oximeters Shift from Overestimation to Underestimation

A recent clinical observation has highlighted a concerning trend: the accuracy of pulse oximeters in a hospital setting has reportedly shifted. Once noted for overestimating blood oxygen levels, particularly in patients with darker skin tones, these devices are now showing instances of underestimation. This reversal complicates clinical interpretation and diagnostic reliability, impacting patient care.

This shift isn't merely a calibration issue; it points to deeper complexities in how medical devices are designed, tested, and deployed across diverse populations. The initial concern stemmed from studies, including one published in the New England Journal of Medicine in 2020, which found that pulse oximeters frequently overestimated oxygen saturation in Black patients compared to white patients when measured by arterial blood gas. This led to delayed interventions and poorer outcomes. The new observation suggests a compensatory adjustment may have overshot, introducing a different kind of error.

The Food and Drug Administration (FDA) has issued warnings and guidance on pulse oximeter accuracy, especially concerning skin pigmentation. In February 2021, the FDA stated, 'Pulse oximeters have limitations and a risk of inaccuracy under certain circumstances.' The challenge lies in ensuring that new algorithms or hardware adjustments don't merely swap one bias for another, but genuinely improve accuracy across all user demographics.

Individuals relying on consumer-grade pulse oximeters, often integrated into smartwatches or fitness trackers, should exercise caution and be aware of these inherent limitations. Prioritize devices that offer transparent data on their validation across diverse populations, and always cross-reference unusual readings with clinical advice rather than relying solely on a single device's output.

One headline rarely tells the story. See how today’s news fits the bigger shifts on AI Trends, or learn to read your own data on How it works.

Keep reading

Based on what you've been reading — always learning.

See all →