Inverse linear association between blood haemoglobin and oxygen saturation accuracy measured by pulse oximetry: a cross-sectional analysis in individuals with COVID-19 infection

Background: Pulse oximetry measures oxygen saturation non-invasively by using differential absorption of infrared signals which are dependent on the oxyhaemoglobin:deoxyhaemoglobin ratio. We tested the hypothesis that pulse oximetry error in measurements of blood oxygen saturations may be associated with blood haemoglobin levels.

Methods: The study design was an observational study of all adult patients admitted to a large teaching hospital with suspected or confirmed COVID-19 infection from February 2020 to December 2021 who had arterial blood gases (ABG) drawn. The pulse oximetry reading was compared with the arterial saturation on the ABG and the measurement error was determined according to the ABG haemoglobin. A secondary analysis was performed among a subset of patients with venous haemoglobins drawn within 24 hours, comparing measurement error between ABG arterial saturation and pulse oximetry readings between those with normal (150 g/L) and low (70 g/L) haemoglobins.

Results: The analysis used 5922 paired oxygen saturations from 3994 patients with contemporaneous haemoglobin measurements by ABG. A 1 g/L decrease in blood haemoglobin was associated with an 0.021% (95% CI: +0.008% to +0.033%) increase in the measurement error (in the direction of a falsely elevated reading.). In the 1086 patients who had had a venous haemoglobin there was a 0.055% (95% CI: +0.020% to +0.090%) increase in the measurement error of oxygen saturation per 1 g/L decrease in blood haemoglobin. The measurement error was thus greater in those with anaemia than in those with normal haemoglobin.

Conclusion: As blood haemoglobin decreases, the oxygen saturation measurement derived from a pulse oximeter reads erroneously higher than the true value measured by ABG. While this study was confined to patients with COVID-19, physicians should be aware of this potential discrepancy among all patients with haemorrhage or known anaemia.