Apple Watch can show a respiratory-rate value in the Vitals app after it is worn to sleep. The number may look like a simple count, but it has several layers: a wearable estimate, a sleep-window context, a personal typical range, and an interpretation.

Keeping those layers separate prevents a common mistake—treating one night’s value as a clinical test or assuming that a change has one confirmed cause. Before comparing nights, preserve the sleep window, device context, source, and missing coverage.

At a glance

What to keep in view

  • HealthKit records respiratory rate as a quantity sample that Apple Watch can record automatically.
  • Vitals uses a personal typical range for overnight metrics; it is not a population normal range.
  • A change can be reviewed with fit, sleep duration, illness, altitude, medication, and coverage context, but it does not prove a cause or diagnosis.
  • A useful AI summary carries timestamps, units, provenance, baseline context, and uncertainty with the value.

What the Vitals app actually shows

Apple lists respiratory rate among the overnight metrics shown in Vitals, alongside heart rate, wrist temperature, blood oxygen, and sleep duration. Apple says the app establishes a typical range for each metric collected while the person sleeps.

That typical range is personal context. It is not a population normal range, and it does not mean that every value inside the range is healthy or every value outside it is dangerous. The purpose is to help someone notice a change relative to their own recent recorded pattern.

  • Check whether the watch was worn long enough and whether sleep tracking was active.
  • Keep the baseline period and the current night visible in an export.
  • Treat possible factors in the Vitals UI as context prompts, not causal conclusions.

What HealthKit stores—and what it does not disclose

HealthKit defines respiratoryRate as a quantity sample using count-per-time units. Apple documents that the system automatically records these samples on Apple Watch. The API tells an app what kind of value it can query; it does not expose the complete estimation algorithm or the uncertainty of an individual sample.

A robust export should preserve the sample’s timestamp, unit, source, device metadata, and any available sleep relationship. Do not export only a rounded value and then imply that it is a direct clinical respiratory examination.

Keep record, estimate, baseline, and interpretation separate

A recorded sample is the quantity saved to HealthKit. An algorithm estimate is the output of the watch’s sensing and processing pipeline. A personal baseline is the Vitals range built from recent overnight records. An interpretation is a human or AI explanation of whether the new value differs and what remains unknown.

This separation matters because sleep duration, wrist fit, movement, altitude, medication, alcohol, illness, and signal gaps can all change the context. A baseline deviation is a reason to inspect the record, not proof of a respiratory condition.

Why validation evidence stays attached to device and setting

Respiratory-rate validation studies compare a wearable estimate with a reference method such as capnography or a ventilator monitor. The result applies to the tested device, algorithm, population, and protocol—not automatically to Apple Watch Vitals.

One peer-reviewed brief report evaluated a wearable photoplethysmography-based monitor in 35 healthy volunteers, 18 ventilated patients, and 92 patients with COVID-19. It used capnography or ventilator monitoring as comparators and reported correlation and Bland–Altman analyses. This is useful methodological context, not Apple-specific accuracy evidence.

  • Record the device and software version whenever a validation study reports them.
  • Check the reference method, setting, sample size, and excluded observations.
  • Do not transfer clinical-monitor agreement limits to a consumer watch without direct evidence.

A synthetic overnight example

The following values are invented and contain no real health data.

  • Monday: 15.2 breaths/min, within the person’s typical range.
  • Tuesday: 17.0 breaths/min, with a short sleep period and a loose-fit note.
  • Wednesday: no sample because the watch was not worn overnight.
  • Defensible summary: Tuesday differed under a different coverage context; Wednesday is unknown.

What an AI analysis should preserve

Before asking a model to summarize respiratory-rate data, include the exact sample time, unit, sleep window, source, device, software version, wear duration, fit warnings, missing intervals, personal typical range, extraction time, and duplicate-source handling.

The model can report a deviation from the recorded personal baseline and list alternative explanations. It should not invent a universal normal range, infer a respiratory diagnosis, or present an Apple estimate as capnography.

MCP may be an optional third-party interface for sending a minimized, user-approved respiratory-rate summary to a compatible AI client. MCP is not an Apple protocol, does not grant HealthKit permission, and does not increase measurement validity.

Privacy and medical boundaries

Respiratory rate, sleep timing, and Vitals notifications can reveal sensitive health information. Request only the respiratory and sleep fields needed for the stated question. Prefer a selected date range and aggregate trend when raw timestamps are unnecessary. Disclose external model processing, retention, deletion, and revocation clearly.

Apple states that Vitals measurements are not intended for medical use. This article does not diagnose breathing disorders, interpret a single value, recommend medication changes, or replace a clinician’s assessment. Persistent symptoms or breathing difficulty require appropriate professional support.

Use boundary

Information, not medical advice

This article explains data and research methods. It does not diagnose a condition, prescribe treatment, establish a universal normal range, or replace qualified professional care. If symptoms or a medical decision concern you, use an appropriate clinical service.