Happy hypoxia in COVID-19: pathophysiology and pulse oximetry accuracy

Many patients with COVID-19 have presented to emergency departments with arterial hypoxaemia but without breathlessness; this is called ‘happy hypoxia’ or, more accurately, ‘silent hypoxaemia’. Hypoxaemia needs to be identified correctly in patients with COVID-19 as it is associated with in-hospital mortality. The aetiology of silent hypoxia is unclear, and the pathophysiological processes involved in the relationship between the response to hypoxaemia and the sensation of dyspnoea may explain its clinical presentation. Pulse oximetry is used routinely to measure oxygen saturation. However, recent literature has questioned its accuracy in patients with COVID-19. Inaccuracies in readings, which arise for several reasons, could in part explain silent hypoxaemia. Caution should be taken when interpreting pulse oximeter readings or patients could be given a higher inspired oxygen fraction than necessary. Silent hypoxaemia may also mask disease severity in patients with COVID-19.

The ongoing coronavirus disease 2019 (COVID-19) pandemic caused many challenges. Among these is the phenomenon of ‘happy hypoxia’ or, more precisely, silent hypoxaemia. Many patients have presented to emergency departments with low oxygen saturations (SpO2) measured with pulse oximetry yet had minimal signs of respiratory distress (Tobin et al, 2020). Although happy hypoxia has gained extensive media coverage, its aetiology remains unclear.

A pathophysiological explanation of silent hypoxaemia in COVID-19 is a dissociation between hypoxaemia and the sensation of breathlessness. The process could involve an idiosyncratic action of COVID-19 on chemoreceptors sensitive to oxygen (Tobin et al, 2020), intrapulmonary shunting, relatively preserved lung compliance, and dysfunctional hypoxaemic vasoconstriction (Dhont et al, 2020). Tobin et al (2020) presented three cases where patients with a reduced partial pressure of arterial oxygen (PaO2) ranging between 4.8kPa and 6kPa denied any having an difficulty in breathing.

Pulse oximetry is a simple, cheap and non-invasive method that has been advocated by many as a ‘potentially life-saving solution’ that allows patients to monitor their ‘oxygen levels’ at home (Gallagher, 2020).

However, recent literature has questioned the accuracy of pulse oximetry in patients with COVID-19, suggesting that happy hypoxia could, in part, be explained by pulse oximeter inaccuracies (Wilson-Baig et al, 2020). Interestingly, in the patients discussed by Tobin et al (2020), the SpO2 ranged from 62% to 76%, while the arterial oxygen saturation (SaO2) ranged from 69% to 83%.

Considering that hypoxaemia in patients with COVID-19 is independently associated with in-hospital mortality (Xie et al, 2020), hypoxaemia must be correctly recognised early in the disease process as this may affect the success of potential management strategies and whether a patient is at risk of requiring admission to intensive care.

This article explores the conundrum of happy hypoxia in COVID-19 and the possible mechanisms involved. A concise description of the physiology of dyspnoea, hypoxaemia and hypoxia is provided, as well as the principles underpinning pulse oximetry.

Physiology of dyspnoea

Dyspnoea is characterised by the sensation of uncomfortable or difficult breathing, breathlessness or experiencing air hunger (American Thoracic Society, 1999).

Coccia et al (2016) describe the aetiology of dyspnoea as physiological, pathological or social. Pathological causes include cardiorespiratory compromise, metabolic imbalances and neuromuscular conditions. The sensation of increased respiratory effort and breathlessness is subjective to the individual, and is caused by a mismatch between the respiratory drive and pulmonary ventilation (Dhont et al, 2020). Essentially, the need for ventilation (afferent signalling) is not met by the physical process of breathing (efferent signalling).

The afferent signalling pathway includes acid-sensing ion channels, mechanoreceptors in the chest wall, vagally mediated stretch receptors found in the smooth muscle of large airways, rapidly adapting stretch receptors (irritant receptors), chemical feedback from peripheral and central chemoreceptors, and juxtacapillary receptors, which are sensitive to pulmonary interstitial oedema.

The efferent pathway consists of descending neuronal motor signalling to the respiratory muscles, of which the diaphragm is the most important. The central processing unit or respiratory centre is located in the medulla oblongata and the pons of the brainstem. The respiratory centre's role is to assess whether efferent motor signalling to the respiratory muscles meets the demands sensed by afferent signalling. This includes airway pressure, airflow and lung movement (Coccia et al, 2016).

Where there is a discrepancy between afferent and efferent signalling—that is, ventilatory muscles are inadequately responding to the command—the sensation of dyspnoea is heightened (Manning and Schwartzstein, 1995). Concurrent stimulation of the sensory cortex results in the conscious sensation of ventilatory effort and dyspnoea (Nishino, 2011). The respiratory centre is also influenced by pain and emotional stimuli through the cerebral cortex, hypothalamic integrative nociception from muscle and lung stretch receptors, and metabolic rate (Dhont et al, 2020). These may increase the individual's sensation of dyspnoea.

Of all the mechanisms that control respiratory drive, the information provided by peripheral and central chemoreceptors is the best-recognised determinant of respiratory drive (Vaporidi et al, 2020). Peripheral chemoreceptors in the carotid bodies are sensitive to arterial oxygen, carbon dioxide (CO2), and pH, while chemoreceptors in the aortic body are sensitive only to arterial oxygen and carbon dioxide. Central chemoreceptors are sensitive to pH changes resulting from arterial changes in carbon dioxide. The ventilatory and dyspnoeic responses to hypoxaemia are mainly influenced by increasing arterial carbon dioxide (Dhont et al, 2020; Vaporidi et al, 2020) as opposed to just hypoxaemia in isolation.

At a steady state, arterial carbon dioxide (PaCO2) is directly proportional to the rate of CO2 production owing to metabolism and inversely proportional to the minute ventilation (litre/minute) and (1 – Vd/Vt) where Vd is the dead space volume (volume of air not participating in gas exchange), and Vt is the tidal volume (Dhont et al, 2020). As PaCO2 rises, the work of breathing is increased by the drive to remove carbon dioxide.

Factors reducing the ability to remove carbon dioxide include increased dead space, reduced lung inflation and higher airway resistance (e.g. because of asthma). Although the usual response to hypoxaemia is a rise in minute ventilation, through increased respiratory rate and tidal volume, hypoxaemia itself plays a limited role in the sensation of experienced dyspnoea (Dhont et al, 2020). This is unlike hypercapnia, where the work of breathing is influenced by the drive to clear carbon dioxide.

Hypoxaemia or hypoxia?

The terms hypoxaemia and hypoxia are commonly used interchangeably. However, these terms are not synonymous. Hypoxaemia is defined as an abnormally low partial pressure of oxygen in the blood when compared to its normal value at sea level (hence the ‘-aemia’ ending). Hypoxia is a term that describes reduced tissue oxygenation—an imbalance between tissue oxygen supply and consumption results in less oxygen being available to maintain cellular function.

Hypoxaemia can occur without hypoxia if there is a compensatory rise in haemoglobin and cardiac output as seen in, for example, a mountaineer at the summit of Everest (Grocott et al, 2009). Similarly, hypoxia can occur without hypoxaemia, where cells cannot use oxygen despite the individual having normal blood and tissue oxygen e.g. in cyanide poisoning (McLellan and Walsh, 2004).

To further understand the pathophysiological concept of ‘silent hypoxaemia’ in COVID-19, hypoxaemia and carbon dioxide clearance should be considered separately. There appears to be a dissociation between the severity of hypoxaemia and respiratory discomfort experienced by patients with COVID-19.


The typical response to mild hypoxaemia is a rise in respiratory rate and tidal volume but not a sensation of dyspnoea. This is because carbon dioxide is not retained and the individual can maintain carbon dioxide clearance. In healthy people, the respiratory drive is minimally altered in mild hypoxaemia (PaO2=8–9kPa), as occurs at higher altitudes (Vaporidi et al, 2020).

However, as the severity of hypoxaemia worsens (PaO2<5.5kPa), dyspnoea often occurs (Manning and Schwartzstein, 1995). Rising respiratory rate and tidal volumes—not the sensation of dyspnoea—are essential clinical signs in assessing for impending hypoxaemic respiratory failure.

The pathophysiological mechanisms by which hypoxaemia occur are summarised in Table 1. Generally, the aetiology of hypoxaemia is owing to one or a combination of reduced alveolar oxygen, reduced blood gas permeability or reduced blood perfusion of gas exchange surfaces (Treacher and Leach, 1998).

Table 1. Causes of hypoxaemia
Table 1. Causes of hypoxaemia
Reduced alveolar oxygen Hypoventilation, high altitude, opiate overdose, neuromuscular disease
Ventilation perfusion mismatch Ventilation-perfusion mismatch (V/Q mismatch) is where blood flow (Q) to an area of the lung is not matched with adequate ventilation (V) or vice versa. This is the most common cause of hypoxaemia. Examples include pneumonia, pulmonary oedema, pulmonary embolism and obstructive sleep apnoea
Reduced oxygen diffusion from alveoli into pulmonary capillaries Interstitial lung disease, acute respiratory distress syndrome (ARDS)
Shunt physiology A shunt is where blood flows from the venous circulation to the arterial circulation without passing through oxygenated alveoli in the lung. A normal physiological response to shunt is hypoxic pulmonary vasoconstriction. This process diverts blood from dysfunctional areas of the lung to better-ventilated areas, thereby optimising V/Q matching (Dunham-Snary et al, 2017). Increasing inhaled oxygen administration does not result in improved oxygen saturation. Pathological causes of a shunt include ARDS, severe pulmonary oedema, lung collapse, atelectasis and right-to-left intracardiac shunt

Carbon dioxide clearance

Increased PaCO2 is the major contributing factor to dyspnoea. The normal response to hypercapnia is an increase in respiratory drive and minute volume ventilation (Dhont et al, 2020), as carbon dioxide clearance is influenced by minute ventilation, the rate of carbon dioxide production and the amount of dead space present.

Scarred alveoli resulting from acute respiratory distress syndrome or a massive pulmonary embolism will reduce carbon dioxide clearance, which will increase the work of breathing through increased tidal volume and respiratory rate.

When the efferent signalling (physical process of breathing) does not match afferent signalling (the need for ventilation), the sensation of dyspnoea is heightened as the individual tries to clear carbon dioxide.

Interestingly, the range of PaCO2 in patients with COVID-19 in the study by Tobin et al (2020) with reduced PaO2 (4.8–6kPa) was 4.5–5.4kPa. The normal range for PaCO2 in a healthy individual is generally around 4.5–6kPa. These individuals did not describe any difficulty with their breathing (Tobin et al, 2020).

Possible pathophysiological explanation for ‘happy hypoxaemia’

Happy hypoxaemia is not exclusive to COVID-19 and can occur in patients with atelectasis, intrapulmonary shunts or right-to-left intracardiac shunt (Dhont et al, 2020). The balance between ventilation and pulmonary capillary blood flow determines the adequacy of gas exchange.

In the initial phase of COVID-19, viral-related inflammation owing to alveolar infiltrates results in a sizeable V/Q mismatch (Archer et al, 2020). This is where blood flow (Q) to an area of the lung is not matched with adequate ventilation (V).

With continued inflammation, lung permeability increases, resulting in progressive oedema and alveolar collapse. Over time, an increasing proportion of the cardiac output is perfusing non-aerated lung tissue. This results in intrapulmonary shunting (Gattinoni et al, 2020).

Also, there is evidence that COVID-19 might inhibit hypoxic pulmonary vasoconstriction, which results in the persistence of pulmonary blood flow to non-aerated lung alveoli (Dhont et al, 2020). In healthy individuals, pulmonary vasoconstriction occurs in response to reduced alveolar ventilation to minimise shunt (Table 1). This process diverts blood from dysfunctional areas of the lung to better ventilated areas, thereby optimising V/Q matching (Dunham-Snary et al, 2017).

In the absence of hypoxic vasoconstriction, red blood cells passing through affected areas of the lung will not get adequately oxygenated because diffusion is impaired. It is unclear whether hypoxic pulmonary vasoconstriction inhibition results from the inflammatory-mediated release of endogenous vasodilator prostaglandins, bradykinin or cytokines (Nagaraj et al, 2017). There is some evidence that COVID-19 mediated mitochondrial damage in the pulmonary artery smooth muscle cells may explain lung perfusion regulation loss (Dhont et al, 2020).

As the infection process progresses, lung diffusion capacity can become impaired. The COVID-19 virus proliferates within alveolar type II cells and resulting in a large viral load. Once released, the viral load is met by the host's immune response, resulting in a virus-linked pyroptosis, which is a highly inflammatory form of programmed cell death (Tay et al, 2020). A loss of alveolar epithelial cells follows, which leads to the denuded alveolar basement membrane being covered with debris consisting of immune response-related activation products, dead cells and fibrin (Tay et al, 2020). The presence of this debris (also known as the hyaline membrane), acts to impair diffusion capacity. In the absence of hypoxic vasoconstriction and a hypercoagulable state, V/Q mismatch worsens, resulting in deteriorating arterial hypoxaemia (Dhont et al, 2020).

Interestingly, in the early stages of COVID-19, airway resistance is maintained. Lung compliance is also normal in patients without pre-existing lung disease, so the work of breathing remains low (Gattinoni et al, 2020). Preserved lung compliance without increased physiological or anatomical dead space suggests that lung gas volume is maintained. These observations could explain why, early in the disease process, patients with COVID-19 do not experience dyspnoea despite being hypoxaemic (Archer et al, 2020; Dhont et al, 2020; Gattinoni et al, 2020).

With continued deterioration, the lung volume available for gas exchange decreases owing to increased consolidation. Lung compliance is reduced, possibly because of lowered surfactant activity, and physiological dead space is increased. V/Q mismatch is also exacerbated because of a hypercoagulable state where microthrombi formation reduces alveolar blood blow (Dhont et al, 2020). These processes increase hypoxaemia, reduce carbon dioxide clearance and increase the work of breathing. With the increased anxiety experienced by patients with COVID-19, the sensation of dyspnoea becomes apparent.

Considering the evidence, the predictors of clinical deterioration in patients with hypoxaemia related to COVID-19 appear to be the rate of increase in respiration rate and tidal volume, not the sensation of dyspnoea.

Pulse oximetry

Pulse oximetry is a simple, cheap and non-invasive method that has been advocated by many as a ‘potentially life-saving solution’ that allows patients to monitor their ‘oxygen levels’ at home (Gallagher, 2021). NHS England was recently involved in an initiative for at-risk patients with COVID-19 symptoms in primary care to monitor their oxygen saturation and relay readings to their health team. Should SpO2 drop, then the patient's doctor can take action (Torjesen, 2020).

Pulse oximetry is based on a combination of optical plethysmography and spectrophotometry. Plethysmography isolates the pulsatile arterial signal by detecting the optical distance changes caused by the arterial pulse. Spectrophotometry is used to identify haemoglobin species by absorbing different light wavelengths (Nitzan et al, 2014; Chan et al, 2013).

The pulse oximeter has two light-emitting diodes that transmit light at two wavelengths—660nm and 940nm, which are red and infrared light wavelengths respectively—and a photodetector that is sited across a tissue bed, such as a finger, toe or ear lobe. The diodes flash around 400 times per second. They are switched on in sequence with a pause when both diodes are off (Ralston et al, 1991; Robertson and Rahemtulla, 2016). This process compensates for ambient light.

The concentrations of deoxyhaemoglobin (HHb) and oxyhaemoglobin (OHb) are determined by the absorption of transmitted light at these wavelengths. HHb absorbs more light at 660nm and OHb absorbs more light at 940nm.

The signal is transduced into a proportional current signal and converted into a voltage. The voltage is amplified, and the interference of ambient light subtracted. The resulting signal is processed to improve the signal:noise ratio and a value displayed on the pulse oximeter. Modern pulse oximeters use complex algorithms to reduce artefacts further.

The change in arterial oxygenation or extra haemoglobin presentation to the sensor, which occurs with every heartbeat, is not related to the pulsatile signal as arterial blood remains uniformly oxygenated with a stable haematocrit throughout each pulsation. What does change is the absorbance of red and infrared light at 660 nm and 940nm respectively. This is related to the optical distance that alters with each pulsation of the arteries. With each beat, the artery expands, resulting in an increased distance between the probe and sensor. The light absorbance, therefore, increases and decreases proportionally with each pulsation. In contrast, the absorbance of red and infrared light at these wavelengths in skin, veins, and bone remains constant as the blood volume does not change with each cardiac cycle.

The Beer-Lambert Law can explain the overall measured absorbance. Essentially, the Beer-Lambert Law is a combination of two laws (Beer's Law and Lambert's Law) that describe the relationship between the absorption of light with the distance the light travels through a substance and its concentration.

Beer's Law states that the amount of absorbed light passing through a solution is directly proportional to the concentration/density of the solution. A beam of light becomes attenuated with increasing concentration/density of a solution. Lambert's Law states that the amount of light absorbed by a substance is proportional to the length of the path the light has to travel through the substance (Simpson et al, 2013).

The Beer-Lambert Law (Figure 1) can be expressed as follows (Nitzan et al, 2014; Chan et al, 2013):

absorbance = log 10 ( Io/I ) = ε cl ⁡


  • Io: incident light intensity
  • I: transmitted light intensity
  • ε: absorption coefficient of haemoglobin. This is a substance-dependent property and is a measure of how strong haemoglobin absorbs light at a particular wavelength of light. In this equation, it is a combination of OHb and HHb coefficients
  • c: haemoglobin (Hb) concentration
  • l: optical path length through blood vessel.

Figure 1. Beer-Lambert Law

Generally, ε and c are constants while l alters with each arterial pulsation. The absorption spectra of arterial blood will therefore change with each heartbeat. The measured absorbance of the variable pulsatile arterial signal is compared to the constant non-pulsatile absorption of tissues and veins at the wavelengths of 660nm and 940nm (Figure 2). This provides the global modulation ratio (R) (Chan et al, 2013) as follows:

R = ( p600/NP660 ) / ( P940/NP940 )


  • R: global modulation ratio
  • P and NP: pulsatile (arterial) and non-pulsatile (tissues and veins) components of the signal at the respective wavelengths.

Figure 2. Measurements and considerations when calculating the global (red/infrared) modulation ratio, which is used to predict arterial oxygen saturation adapted with kind permission from Chan et al. 2013

The global modulation ratio (R value) is used to calculate a corresponding predicted arterial oxygen saturation (Figure 2). The relationship between SaO2 and R is based on the original study by Aoyagi (2003) and studies by Pologe (Pologe, 1987). In these studies, the arterial oxygen saturation was measured against increasing R values. A calibration curve and formula were derived, and SaO2 calculated from the measured R value. The lower the R value, the higher the SaO2, which in turn represents the corresponding SpO2. For example, when R=2, this would give a SpO2 of 50%, while R=0.4 would give a SpO2 of 100%. The percentage value is then displayed as the SpO2 on the pulse oximeter.

It should be noted that the displayed value is an average calculated over a number of beats; therefore, any change in SpO2 will not be displayed immediately but after around 10–15 seconds.

The pulsatile component represents up to 2% of the total absorption (Magee, 2005) and possibly explains of why movement, vasoconstriction or hypothermia may reduce the accuracy of a pulse oximeter.

Limitations of pulse oximetry

In certain circumstances, pulse oximetry does not reliably predict SaO2. Blood gas analysis provides an accurate assessment of oxygenation status and is considered the gold standard.

Sources of error in signal processing result mainly from the calibration process. It is essential to appreciate that pulse oximeters are calibrated by manufacturers using fit, healthy individuals. The original calibration curve was accurately calculated to SpO2 values of 75% (Pologe, 1987). Values below this were extrapolated as it was not ethically possible to subject healthy volunteers to saturations lower than 75%.

The accuracy of pulse oximeters is generally quoted as ±2%. This relates to the standard deviation of differences between SaO2 and SpO2 (Nitzan et al, 2014). However, the availability of various pulse oximeters, including smart watches and modern phones, has raised questions about the accuracy of these devices when measuring SpO2 (Hahnen et al, 2020).

In the hospital setting, Milner and Matthews (2011) examined the accuracy of 847 pulse oximeters used in the NHS. Of these, 10.5% were found to have a functional error in their electrical circuitry and were excluded from the study. Of the remaining 758 pulse oximeter sensors, 22.3% did not conform to the manufacturer's specification. This would have resulted in an inaccurate SpO2 measurement by >4%, which is well outside the ±2% standard manufacturers' claim.

Correct placement of the probe is required to obtain a good assessment of the arterial pulse. Misplacement of the oximeter may result in the ‘penumbra effect’. Here, the paths from the two light-emitting diodes to the photodetector are unequal in length, resulting in the absorption being distorted and one of the wavelengths becoming ‘overloaded’ (Guan et al, 2009). Even after optimising the probe position and using a validated oximeter, unreliable readings of SpO2 can occur in various circumstances (Table 2) (Chan et al, 2013).

Table 2. Mechanisms of and reasons for unreliable pulse oximetry (SpO2) readings
Table 2. Mechanisms of and reasons for unreliable pulse oximetry (SpO2) readings
Inability to read SpO2
  • Poor perfusion owing to hypovolaemia, vasoconstriction, dysrhythmias, peripheral vascular disease, hypothermia
  • Damaged probe
  • Poorly positioned probe
Falsely low SpO2
  • Venous pulsations
  • Excessive movement (especially seen in babies)
  • Shivering
  • Inherited forms of haemoglobin
  • Severe anaemia with hypoxaemia
  • Nail polish
  • Intravenous pigmented dyes used in diagnostic tests
  • Methylene blue
False normal or elevated SpO2
  • Carbon monoxide poisoning
  • Sickle cell anaemia vasoocclusive crises
  • Increased glycosylated haemoglobin in patients with diabetes
Interference with light absorbance
  • Methaemoglobinaemia
  • Sulfhemoglobinaemia
  • Sepsis and septic shock

Peripheral saturation is not an accurate reflection of SaO2 so the latter can be underestimated in low-perfusion states, e.g. dysrhythmias, vasoconstriction, hypothermia, oedema and severe anaemia. Nail polish, oedema and external energy sources (e.g. bright light), can result in erroneous signal measurement. The presence of dyshaemoglobins, such as carboxyhaemoglobin (COHb), methaemoglobin (MetHb) or haemoglobin variants (Sarikonda et al, 2009; Chan et al, 2013). can interfere with absorbance. This is not an issue with modern blood gas analysers as SaO2 is directly measured by spectrophotometric analysis of the haemoglobin released from a sample of haemolysed arterial blood. The incorporated co-oximeter accurately analyses four haemoglobin species (OHb, HHb, COHb and MetHb) as each haemoglobin species has a characteristic light absorption wavelength.

The measured concentration of OHb and HHb allows SaO2 to be calculated as follows:

SaO 2 = ( cOHb ) / ( cOHb + cHHb ) × 100 %


  • cOHb: concentration of oxyhaemoglobin in arterial blood
  • cHHb: concentration of deoxyhaemoglobin in arterial blood.

In critically ill people, SpO2 does not reliably predict equivalent changes in SaO2 (Perkins et al, 2003). This should be expected as the original calibration was based on calculations based on healthy volunteers. Acidosis causes SpO2 to be overestimated in the critically ill (Perkins et al, 2003; Singh et al, 2017) while, in progressive anaemia, SpO2 underestimates SaO2 (Perkins et al, 2003). As SaO2 is the gold standard, this highlights the inaccuracy of pulse oximetry in certain circumstances.

Despite these observations, the clinical significance of these changes is small. Arteriolar dilatation secondary to tissue hypoxia can lead to venous pulsations, which contribute to falsely low SpO2 readings because venous OHb saturation is also measured in the pulsatile vein (Singh et al, 2017) as well as the pulsatile artery.

Other factors, such as increased glycosylated haemoglobin in patients with diabetes, can lead to an SpO2 being overestimatied by pulse oximetry (Pu et al, 2012).

Accuracy of pulse oximetry in COVID-19

The accuracy of pulse oximetry in patients with COVID-19 has been suggested as a potential explanation to the phenomenon of happy hypoxia (Tobin et al, 2020).

Wilson-Baig et al (2020) found that SpO2 did not reliably predict arterial oxygen saturation (SaO2) measured by direct blood gas analysis in 17 ventilated patients with COVID-19 pneumonia. This was despite excluding known causes for SpO2 underestimation (Perkins et al, 2003; Chan et al, 2013; Singh et al, 2017) and confirmation of a good quality trace.

Philip et al, (2020) showed the levels of agreement of SpO2 with SaO2 to be slightly suboptimal from 90 paired samples taken from 30 patients with COVID-19 being stepped down from intensive care onto a general ward or transferred to another intensive care unit.

Several hypotheses explaining this observation have been suggested. First, high ferritin, d-dimer or other proteins in patients with COVID-19 may have altered spectral properties at 660nm and 940nm (Sarikonda et al, 2009). These proteins may negatively impact the signal:noise ratio, thereby reducing the precision of pulse oximetry. Second, COVID-19 may contribute through microvascular complications to tissue hypoxia. This would result in arteriolar dilation and venous pulsations, which would contribute to falsely low SpO2 readings because venous oxyhaemoglobin saturation is also measured in the pulsatile vein (Singh et al, 2017; Chan et al, 2013).

There is anecdotal evidence that anaerobic respiration owing to secondary infection by anaerobic bacteria in patients with COVID-19 might inhibit mitochondrial cytochrome oxidase, thereby causing cellular hypoxia (Chakraborty and Das, unpublished observations, 2020). This mechanism would also result in arteriolar dilation and venous pulsation. Finally, a possible formation of a complex between the virus and haemoglobin may result in increased red-light absorbance relative to infrared absorbance, resulting in a lower SpO2.

Considering that oxygen titration is generally guided by pulse oximetry, evidence suggests that SpO2 should be interpreted with caution in patients with COVID-19. An arterial blood gas assessment may be required depending on the clinical context.


This article has attempted to address the conundrum of happy hypoxia, more precisely called silent hypoxaemia.

In COVID-19, there appears to be a dissociation between the severity of hypoxaemia and respiratory discomfort. In the early stages of the disease, lung mechanics are preserved without increased airway resistance or dead space ventilation. The individual, although hypoxaemic, may not experience breathlessness. With disease progression, sudden and rapid respiratory decompensation may occur. An increased respiratory rate may be the only clinical sign before decompensation. Silent hypoxaemia may therefore mask the severity of the disease in patients with COVID-19.

Pulse oximetry is a widely used, simple, cheap and non-invasive method of measuring SpO2. It has also been advocated for use in primary care to monitor at-risk patients with COVID-19. However, recent literature has questioned the accuracy of pulse oximetry in patients with COVID-19, suggesting that SpO2 does not reliably predict SaO2 in these patients.

The interpretation of SpO2 in a patient with COVID-19 should be taken with caution as inaccuracies in pulse oximetry may result in patients receiving a higher inspired oxygen fraction than necessary. It is unclear why SpO2 underestimates SaO2 in patients with COVID-19. Possible altered spectral properties of high ferritin, d-dimers or other proteins in patients with COVID-19 may impact pulse oximetry precision.

Future research should further assess pulse oximeters' accuracy, especially in the prehospital setting. There are multiple brands of pulse oximeters on the market, the accuracy of which devices is unknown. Further research should also consider how health professionals and patients can interpret the information provided by pulse oximetry together with clinical signs such as increased respiratory rate. Tachypnoea and hyperpnoea could be the most significant warning sign of imminent respiratory failure in COVID-19 patients who do not feel breathlessness.

Key points

  • Patients with COVID-19 may have arterial hypoxaemia although they are not breathless; this silent hypoxaemia is informally known as ‘happy hypoxia’
  • The pathophysiology of silent hypoxaemia in COVID-19 could involve a person having relatively preserved lung compliance and lacking excessive dead space. This means that early in the pathology, CO2 clearance is not an issue
  • Pulse oximetry is a cheap, non-invasive method for measuring oxygen saturation (SpO2). Pulse oximetry works by transilluminating a tissue bed and evaluating the absorption of red and infrared light
  • In certain circumstances, pulse oximetry does not reliably predict arterial oxygen saturation (SaO2). Blood gas analysis is considered to be the gold standard at providing an accurate assessment of SaO2
  • In COVID-19, the accuracy of pulse oximetry is still not understood. High levels of ferritin, d-dimers or other proteins associated with COVID-19 could increase the signal-to-noise ratio thereby reducing the precision of pulse oximetry

CPD Reflection Questions

  • What is your understanding on the physiology of dyspnoea and how would you identify it in your practice?
  • What is your understanding of hypoxia and hypoxaemia, and of how ‘happy hypoxia arises in COVID-19? How would you approach and manage a patient who is not breathless but has an SpO2 of 85%?
  • Consider the mechanism and limitations of pulse oximetry

Keep up to date with Journal of Paramedic Practice!

Sign up to Journal of Paramedic Practice’s regular newsletters and keep up-to-date with the very latest clinical research and CPD we publish each month.