Methodology

Scientific foundation

5HRV uses three well-established short-term heart rate variability metrics, each supported by decades of physiological research.

RMSSD

The square root of the mean of the squares of successive NN-interval differences. RMSSD isolates high-frequency beat-to-beat variation and is widely regarded as the primary time-domain index of cardiac vagal (parasympathetic) modulation. Because it filters out slower trends, it is less confounded by sympathetic or baroreflex influences than other time-domain measures.

SDNN

The standard deviation of all NN intervals during the recording. In a five-minute recording, SDNN reflects total short-term variability, incorporating both sympathetic and parasympathetic influences as well as slower regulatory processes. It is not a pure marker of any single autonomic branch.

LF/HF ratio

The ratio of low-frequency (0.04–0.15 Hz) to high-frequency (0.15–0.4 Hz) spectral power. LF/HF provides information about the relative distribution of spectral power but is not a direct measurement of sympathovagal balance. The LF component contains baroreflex, parasympathetic and other influences in addition to sympathetic activity.

Key references

• Task Force of the ESC/NASPE. Heart rate variability: standards of measurement, physiological interpretation, and clinical use. Circulation 1996; 93(5): 1043–1065.
• Shaffer F, Ginsberg JP. An overview of heart rate variability metrics and norms. Front Public Health 2017; 5: 258.
• Billman GE. The LF/HF ratio does not accurately measure cardiac sympatho-vagal balance. Front Physiol 2013; 4: 26.
• Laborde S, Mosley E, Thayer JF. Heart rate variability and cardiac vagal tone in psychophysiological research. Front Psychol2017; 8: 213.

Reference populations

5HRV uses age- and sex-specific reference percentiles from the DanFunD population study.

The DanFunD cohort

The Danish Study of Functional Disorders (DanFunD) is a large, population-based cohort study conducted in Denmark. The HRV reference data were published by Brinth et al. (online 2022; print version 2024, cited throughout this site).

Percentiles are derived from Table 2 (SDNN, RMSSD) and Table 3 (LF/HF) of the published paper. Reference data cover five age bands and two sex categories, providing 10 separate reference distributions per metric.

The direct 5HRV workflow uses five minutes of supine rest followed by an exact five-minute analysis window. DanFunD recorded seven minutes after the resting period and analysed the final five minutes. The posture, resting period and analysed duration are aligned, but the acquisition workflow is not identical.

5HRV algorithms

Calculation from RR intervals

When Polar H10 measurement is used, 5HRV calculates the HRV parameters directly from the received RR intervals. Processing is performed locally in the browser.

1. RR intervals are converted to milliseconds.
2. Invalid, missing, additional and abruptly abnormal intervals are identified.
3. Missed or additional beat detections are corrected structurally where possible.
4. Remaining isolated abnormal intervals are corrected using cubic-spline interpolation.
5. Recordings with excessive artefacts or substantial signal loss are rejected.
6. Smoothness-priors detrending is applied with λ = 500.
7. RMSSD, SDNN and pNN50 are calculated from the detrended complete NN-interval series.
8. Mean RR and mean heart rate use the corrected, non-detrended intervals.
9. The tachogram is interpolated at 4 Hz.
10. Spectral power is calculated over an exact 300-second sample using a 300-second Welch window.
11. LF is integrated over 0.04–0.15 Hz and HF over 0.15–0.40 Hz.
12. LF/HF is calculated as absolute LF power divided by absolute HF power.

The formulas and principal analysis settings are aligned with published HRV methodology and the settings reported for the DanFunD study. The 5HRV artefact-correction implementation is a transparent custom implementation and is not the proprietary Kubios correction algorithm. Numerical equivalence with Kubios has not yet been established through independent analysis of identical real-world RR recordings.

Percentile interpolation

Individual RMSSD and SDNN values are placed within the 5-point reference distribution using linear interpolation between the published P5, P25, P50, P75 and P95 anchors. The result is a continuous percentile estimate (0–100) that avoids the discontinuities of band-based classification alone. Below P5 and above P95 the estimate is a linear extrapolation rather than an interpolation, and displayed percentiles are limited to the 1st–99th range; such values should be read as approximate.

LF/HF percentiles are computed using logarithmic interpolation, which better reflects the right-skewed distribution of spectral ratio values in the general population. Values below P5 or above P95 are clamped to the 5th or 95th percentiles respectively.

Percentile bands displayed in the report (below P5, P5–P25, P25–P75, P75–P95, above P95) are derived from the same reference distribution using standard percentile cut-points.

Autonomic Pattern Score

The 5HRV Autonomic Pattern Score is a directional composite developed by 5HRV. It combines RMSSD and LF/HF percentiles into a single continuous scale from −100 (parasympathetic direction) to +100 (sympathetic direction). The score provides a research-based summary of the autonomic pattern observed in the recording.

Calculation steps

1. Convert to Z-scores. RMSSD and LF/HF percentiles are transformed to standard normal deviates (Z-scores) using the inverse normal (probit) function. This places both metrics on a common dimensionless scale.
2. Directional sign. The RMSSD Z-score is negated so that low RMSSD (negative Z) contributes positively to the sympathetic-direction score. The LF/HF Z-score is used directly (high LF/HF → positive contribution to sympathetic direction).
3. Weighted combination. The directional deviations are combined with a 70/30 weighting: RMSSD (vagal modulation) contributes 70%, LF/HF (spectral pattern) contributes 30%.
4. Normalization. The combined deviation is divided by the normalization constant 2.2 and scaled to 100. The result is clamped to the range [−100, +100].

Weighting rationale

RMSSD receives the larger weight (70%) because it isolates high-frequency parasympathetic modulation with fewer confounding influences than frequency-domain ratios. LF/HF contributes a smaller weight (30%) as a complementary spectral component while acknowledging its known limitations as a sympathovagal marker.

Pattern classification

The score is classified into six directional categories plus central and mixed patterns. Concordance between the two axes (RMSSD low/high and LF/HF low/high) determines whether the pattern is concordant (both metrics point in the same direction), single-axis (one metric outside central range while the other is within), mixed (both metrics low or both high), or central (both within the central range). A mixed concordance is labelled “Mixed autonomic pattern” regardless of the numeric score.

The concordance classification and single-axis / mixed categories are determined by whether each metric (RMSSD and LF/HF percentile) falls below P25, above P75, or within the central P25–P75 range. Concordance categories are displayed in the report alongside the numeric score and pattern label.

Scope

5HRV is a scientific HRV calculator. It analyses five-minute HRV recordings using peer-reviewed physiological research and published age- and sex-specific reference populations. The output describes the recorded HRV data in relation to these references — it is not a clinical assessment of the individual.

Clinical interpretation remains the responsibility of the healthcare professional, who integrates the HRV findings with the patient's history, symptoms, examination, medications, recording conditions and other relevant clinical information.

Limitations

• Recording context. HRV is sensitive to body position, breathing rate and depth, time of day, prior activity, caffeine, nicotine, food intake and medications. Standardized recording conditions are required for meaningful interpretation.
• Five-minute recordings. Short-term HRV captures resting autonomic modulation but does not reflect circadian variation, sleep-stage architecture, or responses to daily stressors. Five-minute values are not interchangeable with 24-hour Holter metrics.
• Recording quality. Ectopic beats, atrial fibrillation, frequent supraventricular or ventricular ectopy and movement artefacts can invalidate HRV analysis. The calculator assumes technically valid NN-interval data.
• Reference population. The DanFunD cohort reflects a Danish general population aged 18–72. Reference percentiles may not generalise to all ethnic groups, clinical populations, or individuals outside this age range. Above age 72, no age-specific reference percentiles are applied.
• LF/HF interpretation. The LF/HF ratio does not directly measure sympathetic activity or sympathovagal balance. The LF component can contain baroreflex, respiratory and parasympathetic influences. It is used in the score as a directional indicator alongside RMSSD, not as an independent diagnostic marker.
• Autonomic Pattern Score. The 5HRV Autonomic Pattern Score is a composite metric developed by 5HRV. It is intended for research-based analysis of HRV recordings. It has not been independently validated as a diagnostic scale and should not be used in isolation for clinical decision-making.
• Single recording. A single five-minute recording provides a snapshot of autonomic state at one point in time. Serial measurements under standardised conditions are more informative than a single recording. Observed patterns should not be interpreted as chronic or permanent physiological states.