# Accuracy and Bias of Pulse Oximetry in the Intensive Care Unit: A Prospective Observational Study

## TL;DR

This study finds that pulse oximeters in ICU settings are more affected by patient age and oxygen levels than skin tone, highlighting the need for better device accuracy.

## Contribution

The study provides new evidence on skin tone-related bias in pulse oximetry using objective metrics and shows age predicts measurement errors.

## Key findings

- Two of three oximeters exceeded FDA accuracy thresholds in ICU patients.
- Older age was associated with greater oximeter overestimation of blood oxygen levels.
- Skin pigmentation effects were indistinguishable from random error in measurements.

## Abstract

Non‐invasive oximetry is crucial for the easy and continuous monitoring of blood oxygenation, from routine assessments to critical care in intensive care units (ICU). Studies indicate that pulse oximeter readings may be biased due to the high light absorption of melanin.

This study aims to analyse the agreement among three different pulse oximeters and their concordance with arterial blood gas measurements in ICU patients and verify how accuracy varies with age and skin pigmentation.

We conducted a prospective cross‐sectional observational study with a population of intensive care patients. Blood oxygen saturation was measured simultaneously by pulse oximetry and arterial blood gas analysis. Skin pigmentation was objectively assessed at two anatomical sites using calibrated colorimeters and quantified by the individual typology angle (ITA°).

Among 100 ICU patients (SaO2 85%–100%), two of the three oximeters exceeded the 3% accuracy root mean square (ARMS) 2013 FDA threshold, whereas one device remained within limits (2.91%). Agreement with arterial oxygen saturation was weak (CCC = 0.34–0.46) and lower SaO2 was associated with overestimation increase (0.31%–0.54% for every 1% SaO2 decrease). Multivariable models showed no evidence of association between ITA° and SpO2 error (p > 0.05). In contrast, older age predicted greater overestimation (β range = 0.05%–0.06% per year; 95% CI range = 0.02–0.09; all p ≤ 0.05).

Although estimates suggested small pigmentation effects, analysis indicated these were indistinguishable from random error. In contrast, physiological factors such as lower SaO2 and older age were consistently associated with greater oximeter error. Substantial clinical variability limits detection of small biases, underscoring the need for improved calibration and validation using objective skin‐tone assessment.

In intensive care, where continuous monitoring is essential, pulse oximeter error was more strongly related to physiological factors, such as arterial oxygen saturation and age, than to skin tone. These findings support cautious interpretation of SpO2 values and appropriate device selection to ensure accurate assessment across diverse populations.

What is known about the topic?
○Pulse oximetry is an essential tool of non‐invasive monitoring in critical care, yet accuracy can be compromised by device design and patient factors.○Melanin absorption has been implicated in measurement bias, but a lot of studies have relied on self‐reported race rather than objective pigment metrics.○Changes in perfusion and haemodynamics may further degrade signal quality, potentially confounding oximetry readings in ICU patients.
What this paper adds
○Commercial pulse oximeters commonly used in home and clinical settings underperformed against regulatory ARMS thresholds in ICU patients, emphasising the need for rigorous device validation and informed selection.○The study provides new evidence on skin tone‐related bias in pulse oximetry across a diverse patient population, highlighting equity considerations in device performance.○Patient age consistently predicts SpO2–SaO2 discrepancy, underscoring the need to account for perfusion‐related confounders in clinical interpretation.

What is known about the topic?
○Pulse oximetry is an essential tool of non‐invasive monitoring in critical care, yet accuracy can be compromised by device design and patient factors.○Melanin absorption has been implicated in measurement bias, but a lot of studies have relied on self‐reported race rather than objective pigment metrics.○Changes in perfusion and haemodynamics may further degrade signal quality, potentially confounding oximetry readings in ICU patients.

Pulse oximetry is an essential tool of non‐invasive monitoring in critical care, yet accuracy can be compromised by device design and patient factors.

Melanin absorption has been implicated in measurement bias, but a lot of studies have relied on self‐reported race rather than objective pigment metrics.

Changes in perfusion and haemodynamics may further degrade signal quality, potentially confounding oximetry readings in ICU patients.

What this paper adds
○Commercial pulse oximeters commonly used in home and clinical settings underperformed against regulatory ARMS thresholds in ICU patients, emphasising the need for rigorous device validation and informed selection.○The study provides new evidence on skin tone‐related bias in pulse oximetry across a diverse patient population, highlighting equity considerations in device performance.○Patient age consistently predicts SpO2–SaO2 discrepancy, underscoring the need to account for perfusion‐related confounders in clinical interpretation.

Commercial pulse oximeters commonly used in home and clinical settings underperformed against regulatory ARMS thresholds in ICU patients, emphasising the need for rigorous device validation and informed selection.

The study provides new evidence on skin tone‐related bias in pulse oximetry across a diverse patient population, highlighting equity considerations in device performance.

Patient age consistently predicts SpO2–SaO2 discrepancy, underscoring the need to account for perfusion‐related confounders in clinical interpretation.