If two patients receive the same oxygen reading, does that mean the instrument measured them equally well? Averages can make a group look tidy while concealing meaningful differences among the people in it. The physics of a measurement sets limits, but so do the bodies and conditions through which a signal travels. Looking at individual variation helps reveal what a summary number can and cannot tell us.
One group, many readings
Imagine measuring oxygen saturation in a room of patients, then calculating the mean. That average is a useful description of the group, but it does not say whether most readings clustered tightly around it or whether some were far higher and lower. Nor does it tell us who those patients were. A mean is made by adding values and dividing by their number; in that operation, the identity and circumstances behind each value disappear.
This matters because instruments do not encounter an abstract average person. A fingertip pulse oximeter sends red and near-infrared light through tissue and detects how oxygenated and deoxygenated blood absorb those wavelengths differently. The device estimates oxygen saturation from the resulting signal. Blood flow, motion, nail polish, and skin pigmentation can all affect that optical path and therefore the reading [1].
A signal passes through a body
Think of shining a flashlight through two different curtains. Even if the light source is identical, the amount reaching the other side depends on the fabric. In pulse oximetry, melanin absorbs some red and near-infrared light. The NHLBI describes evidence that readings can overestimate oxygen saturation in patients with darker skin tones, an error that can matter when a low oxygen level should prompt attention [1]. The device’s calculation can be consistent and still be biased by how the signal interacts with tissue.
Other conditions alter the measurement in different ways. Poor fingertip circulation can weaken the signal; movement can introduce motion artifact, meaning unwanted variation that the device may mistake for a real pulse. Accuracy also tends to decline at lower oxygen saturation, precisely where a reliable estimate is especially important [1]. These are not interchangeable sources of error, and an overall average may conceal that a subgroup experiences one of them more often.
Why trial averages need context
The same reasoning applies to treatment research. A clinical trial may compare average outcomes between groups, but the result depends on who enrolled, what was measured, and how much individuals varied. A trial is a research study designed to test medical approaches in people; different trial types answer different questions, such as whether a treatment works or whether a screening method finds disease earlier [2]. A favorable average in one study therefore does not guarantee an identical result for every participant or every future patient.
Suppose a treatment improves a group’s average measurement by a small amount. That could reflect a modest improvement for nearly everyone, a large improvement for a few, or a mixture of benefit and no change. The mean alone cannot distinguish these patterns. Researchers use additional information—such as the spread of results, the number of participants, and how outcomes differ across subgroups—to understand how securely the finding applies beyond the study.
The useful question behind the number
Averages remain valuable: they make comparisons possible and can show patterns that individual anecdotes cannot. But interpretation improves when we ask what was averaged and what differences might have been hidden. For a measuring device, that means considering signal quality and the people whose bodies produce the signal. For a trial, it means looking at the design and participants as well as the headline outcome [2].
That approach is especially important when measurements guide decisions. During the COVID-19 pandemic, concerns about pulse-oximeter overestimation and delayed intervention highlighted how device limitations can compound existing health disparities [1]. The lesson is not that one summary is useless. It is that the group statistic is a starting point: the spread around it, the measurement process, and the patients represented determine what it can responsibly mean.
Sources & further reading
- Updating Pulse Oximeters NHLBI
- Clinical Trials: What Are the Phases? National Cancer Institute
Prepared for the independent PSI Center reading library. About edition dates and source checking.
This explainer describes scientific principles. It does not assess individual treatment choices.
