Why Median and Mean Respiratory Rate Should Not Share One Trend Line

One line can hide two different summaries

A nightly respiratory-rate trend may appear as one value, but the summary depends on how readings were combined. A mean adds the readings and divides by their count. A median sorts them and selects the middle value, or the midpoint of two central values when the count is even. Both describe a center, but they do not always produce the same number.

That difference matters when the readings vary, contain a few unusual estimates, or cover an uneven part of the night. A mean can move toward a high or low value because every reading contributes its distance from the others. A median is less affected by a small number of extreme values, but it also does not describe how spread out the readings are. If a chart changes between these methods without saying so, the apparent trend may reflect a calculation change rather than a change in breathing.

I would want a health interface to show the summary method beside the metric, not bury it in a help page. A statistical definition of mean and median makes clear that these statistics answer related but distinct questions about a collection of values. A reader needs the name of the calculation to interpret what a nightly line represents.

What a mean emphasizes

The arithmetic mean uses every included reading. If most estimates cluster in a narrow band while a few are much higher, those higher values pull the mean upward. The shift may be appropriate if the goal is to reflect total magnitude across the full set. It may be misleading if those few readings came from motion, a poor sensor fit, or an unstable interval that should be reviewed separately.

Potential outliers should be examined rather than automatically removed, as the statistical guidance on unusual observations explains.

The mean is also sensitive to how many readings were collected at different moments. If a device samples more often during one part of the night, that segment can have more influence unless the method first bins or weights the data. A user cannot tell from the plotted line whether every sample counts equally, whether each minute contributes one summary, or whether the algorithm excludes poor-quality periods.

For that reason, an app should state the aggregation window and input rules. Does the nightly mean include all detected sleep, only stable intervals, or only times when the sensor passes a quality threshold? Are missing intervals omitted, interpolated, or left as gaps? If the method is not visible, a precise decimal value can imply more certainty than the underlying sampling supports.

The wearable validation study for nocturnal respiratory rate evaluates estimates against a reference measurement under a defined study method. That kind of evidence can help explain how a particular device performed in the studied conditions. It does not tell a reader whether a specific app’s nightly line is a mean or a median unless the app documents that calculation.

A separate multi-signal sleep-breathing study evaluates wearable respiratory and oxygen inputs against polysomnography under its own methods.

What a median emphasizes

The median identifies the point in the ordered set with half of the included values below and half above it. Because its location depends on rank rather than distance, a few unusually high estimates usually have less effect on the result than they would on the mean. This can make the median a useful summary when a distribution is skewed or when isolated values need to be prevented from dominating a central estimate.

That property has limits. The median can hide brief but meaningful variation. If respiratory rate changes during a part of the night, the median may remain close to the center of the readings and not show when or for how long the change occurred. A median is not a signal-quality filter, a clinical judgment, or proof that an unusual reading was wrong. It is a different summary of the values that were included.

Potential outliers are observations to investigate, not automatically discard. That is a sound approach for wearable data as well. A suspicious point might be a sensor artifact, but it might also be a real event or a different condition in the measurement. The app should identify its handling rule and preserve the ability to inspect the time series when a summary deserves a second look.

When a product reports a median, it should still show enough context to make that value interpretable. The user may need the number of valid readings, the percentage of the sleep interval covered, the time range, and a view of the distribution or variation. A single median without those details can look complete even if large portions of the night were not measured.

The measurement window changes the meaning

Respiratory rate is not a single property observed continuously in every wearable. The device estimates it from available signals during supported conditions. The method may use optical pulse data, motion, or a combination; the specific inputs and quality controls vary by product. A nightly summary therefore combines two questions: what values were estimated, and which parts of the night were eligible to contribute.

An average over the full sleep window can differ from an average of shorter intervals. If the app first computes a value for each segment and then averages those segment summaries, its result may not equal a mean of every raw estimate. If it takes the median of segment medians, the result may differ from the median of all individual readings. These are ordinary consequences of combining data at different levels, but the product should describe its approach in plain language.

The sample example also shows why units and labels matter. Respiratory rate is usually expressed in breaths per minute, but a value alone says little without the interval and the source of the estimate. A nightly summary is not a direct observation of every breath. It may be a model output derived from signals the device can measure while it is worn.

The sleep-medicine position on consumer sleep technology advises that consumer-generated sleep information belongs in the context of a wider evaluation and does not replace medical assessment. That boundary applies to both means and medians. Changing the aggregation rule does not turn a wearable estimate into a diagnosis.

Keep a calculation consistent across nights

If an app changes from mean to median, changes which samples qualify, or recalculates older nights, it should identify the date and explain whether past results were updated. A continuous line should not imply that old and new values share one definition when they do not.

How to review a nightly respiratory-rate value

Start with the label. Look for “mean,” “median,” “average,” or another stated summary method. If the app uses “average” without defining it, ask whether it means the arithmetic mean or a broader summary. Confirm that the unit is breaths per minute and identify whether the time window covers the whole detected sleep interval or only valid measurements.

Check for gaps, non-wear, low-confidence sections, and the number of contributing intervals. Compare nights only when fit, settings, and calculation method are reasonably consistent.

Then look at the range or trace when available. The mean and median together can reveal whether a few values sit far from the center. A chart can show whether those points are clustered at one time or spread across the night. Neither display proves a cause, but both can help determine whether the single summary is concealing an important pattern.

If the value is unexpected, repeat the review across more than one night and check the device’s fit and instructions. Do not interpret a consumer respiratory-rate result by itself as evidence of a sleep disorder. Persistent breathing concerns, symptoms, or a change that worries you should be discussed with a qualified clinician. The general-wellness guidance for low-risk devices supports a clear boundary between wellness information and clinical diagnosis or treatment.

What a clear interface should explain

A useful respiratory-rate trend should identify the calculation, the unit, the time period, and the amount of usable data. It should explain whether high or low readings were excluded, and why. It should mark device or software changes that affect comparisons. Where practical, it should let the user inspect a trace or range so the summary does not stand alone.

If the display uses a personalized baseline, the app should explain how many nights contributed and whether the baseline uses the same summary method as the current value. A baseline built from means should not be compared silently with a median for the current night. A change in data method can create a difference even when the underlying pattern has not changed.

The app should also use cautious language. A higher or lower result describes the device estimate; it does not establish a cause or diagnosis. For broader context, a guide to reading breathing quality and respiratory rate separately explains why two related labels can still represent different features.

Keep the summary matched to the data

Mean and median describe a collection differently. The mean uses every included value; the median is less affected by a few extremes but can hide when variation occurred.

Read either value alongside its method, coverage, and time window. If those details are missing, treat the line as an incomplete summary, not a record of every moment of sleep.

FAQ

Is median respiratory rate more accurate than mean respiratory rate?

Neither summary is always more accurate. The appropriate choice depends on the intended question, data distribution, measurement quality, and whether the method has been validated for that use.

Why can the mean be higher than the median?

A few higher readings can pull the mean upward because the mean uses the magnitude of every value. The median depends on the ordered middle of the set and is less affected by a small number of distant observations.

Can a wearable respiratory-rate trend diagnose a sleep disorder?

No. A consumer trend is an estimate from a device method and should not replace clinical evaluation or diagnostic testing when a health concern is present.