Heart Rate Variability (RMSSD)

Overview

Heart rate variability (HRV) is the beat-to-beat variation in the time between heartbeats. RMSSD, the root mean square of successive differences, is the standard time-domain measure of its short-term, high-frequency component. That component is driven mainly by the vagus nerve (parasympathetic, "rest and digest") and by breathing (respiratory sinus arrhythmia). Shaffer & Ginsberg (2017) call RMSSD "the primary time-domain measure used to estimate the vagally mediated changes reflected in HRV". It is also the HRV measure most recovery-focused wearables report.

HealthKit added this type in iOS 27. Until then, SDNN (HKQuantityTypeIdentifierHeartRateVariabilitySDNN) was the only HRV type. The Health app calls this one Recovery HRV. In HealthKit's own iOS 27 description it is the variation between heartbeats measured "using the RMSSD formula", which is "commonly used to monitor day-to-day recovery and, when combined with other information, estimate readiness". In the same strings, SDNN is described as the overall variation between heartbeats, "better for understanding your underlying, long-term health". Apple's watchOS 27 announcements present the pair as Recovery HRV (daily signals of stress and recovery) and overall HRV (broader health, including cardiovascular health).

How It's Measured

The calculation:

• Start from normal-to-normal (NN) intervals: the times between consecutive normal beats, with ectopic beats and artifacts removed
• Take each successive difference: NN(i+1) − NN(i)
• Square the differences, average them, and take the square root: RMSSD = √( Σ (NN(i+1) − NN(i))² / (N − 1) )
• Because it uses only adjacent beats, slow trends in heart rate (posture changes, circadian rhythm, activity) largely cancel out. What remains is fast, beat-to-beat modulation.
• It carries the same information as the Poincaré-plot measure SD1 (SD1 ≈ RMSSD/√2)

Why short windows are enough (unlike SDNN):

• SDNN measures all variability in a recording, so it grows as the window lengthens and takes in slower rhythms. Short-term and 24-hour SDNN are not comparable.
• RMSSD reaches a stable value quickly. In 3,387 adults, a single 10-second recording gave a valid RMSSD; 120 s agreed almost perfectly with 4-5 minutes; and RMSSD beat SDNN at every recording length (Munoz et al. 2015).
• In a pulse-sensor (PPG) study of 467 people, about 30 s was needed to estimate 5-minute RMSSD, versus about 240 s for SDNN (Baek et al. 2015, as reviewed by Shaffer & Ginsberg 2017)
• This is what makes frequent, short wrist measurements practical for RMSSD

Apple Watch measurement (what Apple has and hasn't said):

• Recorded by Apple Watch Series 12 and Ultra 4 on watchOS 27. Apple says the feature needs these models, so there is no Apple Watch Recovery HRV history from before watchOS 27.
• Apple says HRV is measured "as often as every five minutes", 24 times more often than before. Results appear in a new section of the Heart Rate app and in Vitals, which now has a daytime view.
• Overnight Vitals analyses Recovery HRV against the user's personal baseline
• Not documented: how many seconds each sample covers, the sensor signal and beat-detection method, artifact handling, motion gating, and whether a five-minute cadence holds through the day or only during sleep and rest. Apple's developer documentation for the identifier is still a bare declaration. Don't assume a fixed number of samples per day.
• Apple's own Health description says measuring during sleep "tends to more accurately reflect the level of strain on your body because you're not active"

RMSSD vs. SDNN

| | RMSSD (this type) | SDNN | |---|---|---| | Health app name (iOS 27) | Recovery HRV | Heart Rate Variability | | Measures | Beat-to-beat (short-term) variability | Total variability over the recording | | Main physiological driver | Vagal (parasympathetic) activity, breathing | Both autonomic branches plus slower rhythms | | Sensitivity to window length | Low: valid from very short recordings | High: values grow with recording length | | Typical use | Day-to-day recovery, stress, readiness | Long-term health, clinical risk (24-hour SDNN) | | HealthKit availability | iOS 27 / watchOS 27 | iOS 11 / watchOS 4 |

The two are correlated but not interchangeable, and they have different scales. A short-term RMSSD is usually smaller than an SDNN taken over the same window. Never chart them on one axis or substitute one for the other in a trend.

Health Significance

• Vagal activity: RMSSD is more influenced by the parasympathetic system than SDNN is (Shaffer & Ginsberg 2017). Higher resting values generally mean stronger vagal modulation of the heart.
• Stress and recovery: Acute stressors such as hard training, poor or short sleep, alcohol, illness or psychological stress typically lower it. Apple puts it this way: higher Recovery HRV "usually means your body is more adaptable and ready to respond", while lower values can mean the body "is under stress or tired".
• Training monitoring: Athlete-monitoring research tracks the natural log of RMSSD (lnRMSSD) with averaging over several days rather than single readings, and reads the trend against the athlete's own history (Plews et al. 2013)
• Ageing: RMSSD falls steeply through early and middle adulthood. In 24-hour recordings it dropped to about 47% of its second-decade value by the sixth decade, then levelled off (Umetani et al. 1998). Some studies see it rise again after about 70 (Almeida-Santos et al., via Shaffer & Ginsberg 2017).
• Clinical risk: The 1996 Task Force's risk-stratification evidence centres on 24-hour measures of overall HRV, such as SDNN and the HRV triangular index. Wrist-measured RMSSD is a wellness and trend metric, not a diagnostic test.

Clinical Interpretation Guidelines

Normal Values

Values depend on:

1. Age (see the clinical ranges above)
2. Measurement conditions: sleep, lying, sitting or standing; paced or free breathing
3. Time of day (overnight values usually differ from daytime ones)
4. Heart rate: slower heart rates leave more room for intervals to vary ("cycle-length dependence")
5. Fitness and training state

Reference points:

• Short-term resting recordings in healthy adults: 42 ± 15 ms pooled across studies (Nunan et al. 2010)
• By decade in a population sample (Voss et al. 2015): about 40-43 ms at 25-34, about 32-35 ms at 35-44, about 23-26 ms at 45-54, about 20-21 ms at 55-64, about 19 ms at 65-74. The spread within each decade is large (SD 11-23 ms).
• 24-hour Holter: 27 ± 12 ms (Task Force 1996). This is not comparable with wearable readings.
• None of these were measured with an Apple Watch. Apple publishes no population norms for Recovery HRV and judges it against the user's own baseline instead.

Lower Values (Reduced RMSSD) May Reflect

Acute or transient:

• Short or poor-quality sleep
• Alcohol, especially in the evening
• Acute illness or fever
• Heavy or unaccustomed training load
• Psychological stress
• Travel to high altitude

Chronic:

• Normal ageing
• Low cardiorespiratory fitness
• Conditions associated with reduced vagal HRV, such as diabetes with autonomic neuropathy, heart failure and other cardiovascular disease
• Some medications (Apple notes that certain medications can affect it)

Higher Values (Elevated RMSSD) Generally Indicate

• Good recovery and parasympathetic predominance at rest
• Aerobic fitness (often higher in endurance-trained people)
• Caution: A sudden jump, or a very high value with an irregular rhythm, can be a measurement or rhythm problem rather than good news. Ectopic beats and arrhythmias create large successive differences and inflate RMSSD.

Red Flags for Consultation

• A sustained fall from personal baseline over several days together with a rising resting heart rate, fever or new symptoms
• Persistently low values with symptoms that suggest autonomic dysfunction, such as lightheadedness on standing, fainting or exercise intolerance
• Erratic or unusually high values together with Irregular Rhythm notifications, AFib History, or palpitations. The rhythm needs evaluating, and RMSSD cannot be read as vagal activity until it is.
• A downward trend over weeks or months that lifestyle factors don't explain

Caveats & Limitations

Measurement Limitations

• Artifacts and ectopic beats: Successive differences amplify single mis-detected or premature beats. Even two artifacts in a 5-minute segment can significantly distort RMSSD (Shaffer & Ginsberg 2017).
• Optical sensing: Pulse intervals from PPG are not ECG RR intervals. At rest the agreement is generally adequate, though short-term variability tends to be somewhat overestimated, and movement degrades it (Schäfer & Vagedes 2013).
• Breathing: Changes in breathing rate and depth can change RMSSD without changing vagal tone (Shaffer & Ginsberg 2017)
• Undocumented algorithm: Apple has not published how each sample is computed, so values may not match RMSSD computed by another device, or from raw beat-to-beat data, over the same period
• No back-history: Apple Watch samples start with watchOS 27 on Series 12 and Ultra 4. Existing SDNN history cannot be converted to RMSSD. Where an app has beat-to-beat heartbeat series samples (HKHeartbeatSeriesSample), it can compute RMSSD from them directly.

Interpretation Caveats

• Personal baseline first: The spread between healthy people is wide, so one person's value means more against their own history than against a population norm
• Overnight and daytime are different measurements: Daytime samples carry posture, activity, caffeine and mental load. Don't average them with overnight readings and call the result a recovery score.
• Skewed distribution: RMSSD is right-skewed. Research usually analyses ln(RMSSD), and percentage changes are easier to compare than raw milliseconds.
• Older adults: In Umetani et al. (1998), 12% of healthy people over 65 had 24-hour rMSSD below published mortality-risk cut-points, so a low value alone is not diagnostic

What RMSSD Cannot Tell You

• A diagnosis of any specific disease
• Whether a rhythm is normal (it assumes one)
• Sympathetic ("fight or flight") activity: RMSSD mainly indexes the parasympathetic side
• Blood pressure, glucose or structural heart disease

Additional Notes

For health consultants and coaches:

• Track a rolling multi-day average (often 7 days) of overnight values against the person's own baseline, not single readings
• A drop that persists for several days, together with a higher resting heart rate, is a more useful signal than one low night
• Expect values to fall with age; compare people only with themselves

For developers:

• The identifier exists only in the iOS 27 SDK. Gate it with #available(iOS 27, watchOS 27, *). If you also build with an older Xcode, guard the symbol at compile time too, or spell the identifier string out.
• Discrete (arithmetic) aggregation, in ms. Daily statistics average every sample, so a full-day mean mixes sleep and waking readings. For recovery use, restrict queries to the sleep window.
• Third-party apps can write this type, so filter by source when you need Apple Watch readings only (separateBySource)
• Keep it separate from SDNN in storage, charts and analysis

Comparing devices:

• Oura, WHOOP, Garmin and other RMSSD-based devices use different windows (for example whole-night averages versus short readings), so their numbers are not directly comparable with Apple Watch Recovery HRV or with each other
• Compare values from the same device over time