Oxygen Saturation (SpO2): Normal Values, Pulse Oximeters and Smartwatches — How Accurate Is the Reading?

There is no single number that fits every situation. In healthy adults at sea level saturation is usually around 95-100%, and single brief readings a point or two lower do not have to mean anything. In people with chronic lung disease such as COPD, target values are set by the doctor individually and may be lower, because in some patients too much oxygen can be harmful. That is one reason we do not give rigid thresholds for self-use.

Data from hospital practice do, however, allow some numbers to be placed in context. In a study in hospitals in Australia and New Zealand, hypoxaemia in arterial blood was defined as SaO2 below 90%, and a pulse oximeter value of 92% or higher never accompanied such hypoxaemia. That does not mean 91% is a catastrophe and 92% a safe point for everyone. It does show why clinicians use a safety margin: measurement error can be several points, so a result near the boundary must be interpreted with care.

A conclusion for the reader: a medical pulse oximeter is usually reliable and the average error is about 1 percentage point, but in a few percent of people the reading can be off by 4 points or more. A single odd result is therefore worth repeating, and health decisions should take symptoms into account, not just the number. The study concerns hospital and outpatient patients, not healthy people measuring themselves at home on a random device.

A pulse oximeter works well when light passes through well-perfused, still tissue. Any departure from that can falsify the reading. A review of 23 studies tried to put these factors in order.

A simple practical rule follows: dark polish (black, brown) is worth removing or choose another finger, while with clear or light polish the measurement usually still works. Cold hands are the most common everyday cause of odd, low readings, especially in people prone to cold fingers (we write about Raynaud's phenomenon in our article on cold and whitening fingers). If a low value appears only on a cold finger and disappears when the hand warms up, it is probably a measurement error, not hypoxaemia.

For several years the medical literature has contained evidence that pulse oximeters may overestimate saturation in people with darker skin. This translates into the possibility of missing real hypoxaemia (so-called occult hypoxaemia). The issue matters for anyone using a watch with SpO2 measurement, too.

How to read this? First, the data concern hospital patients, so transferring them to healthy people measuring themselves at home is uncertain; the authors explicitly note the lack of evidence from outpatient settings. Second, the point is not that a pulse oximeter 'does not work' in people with darker skin, but that the reading may be too high and low saturation may go unseen. Third, the practical conclusion is the same for everyone: the number must not be judged apart from how the person feels. Someone with breathlessness, blue colour or confusion whose pulse oximeter shows 96% should not be reassured by the result.

Many watches and bands can estimate SpO2, usually from the wrist, at rest or at night. The result is convenient, but the question of how much it is worth is legitimate: the wrist is a harder place to measure than the finger, and the signal is more sensitive to movement and sensor position. We cite two laboratory studies that compared device readings with measurement in arterial blood.

How do these fit together? At normal saturation and small deviations, which is where the great majority of users measure themselves every day, an error of 2-3 points is acceptable for approximate trend monitoring. It is worse in the range that matters most: where saturation really is low, devices, including the watch, were less accurate and tended to overestimate. So a watch can be a curiosity, but it should not be a reason for reassurance when there are symptoms or for self-diagnosing sleep apnoea or lung disease. Similar limitations apply to VO2 max figures from the wrist, which we discuss in a separate article.

Conversely, a single low reading in a healthy person without symptoms, especially on a cold finger or with polish, most often turns out to be a measurement error. Neither a pulse oximeter nor a watch should cause anxiety without symptoms, or false certainty when they are present.

The pulse oximeter is a useful but limited tool. It is good at excluding serious hypoxaemia when the result is clearly normal, yet its error, dependent on finger temperature, nail polish, movement and skin tone, is larger than commonly assumed. Watches are even less accurate exactly where it matters most, that is at low saturation. For a healthy person it is sensible to treat such readings as a curiosity, not a basis for decisions. When symptoms appear, diagnosis is decided by medical examination and, if necessary, a blood gas test, not by a number on a screen.