pnn20

Proportion Of Successive Differences > 20 Ms

Definition

Proportion of successive differences > 20 ms. Formally: sum(|diff(IBI)| > 20) / N.

What does normal look like?

Fitted normative prior: Beta(α = 1.806, β = 3.301), KS p = 1.9e-17, n = 61715.

Empirical distribution over the pooled nsrdb+nsr2db normative windows (360-beat windows, 120-beat stride), overlaid with the fitted Beta prior density. Vertical lines mark the median and the 5–95% range.

Normal-range summary (pooled nsrdb+nsr2db)

n varies by feature only through per-window validity over the full pooled nsrdb+nsr2db table (n up to 61 715; e.g. sampen/mse drop windows where the statistic is undefined). ulf is the one exception: a 360-beat (~5 min) window contains no ULF-band power, so it uses a long-window NSRDB-only extraction (see its own page).

Use cases

• More sensitive successive-difference count for low-variability or short records.
• Vagal-tone screening where pNN50 saturates near zero.
• Part of the generalised pNNx family (Mietus et al. 2002).

Applications by area

Evidence is reported at the measure-family level; a specific variant may not be the exact index measured in every cited study.

Clinical

Coverage: statistics. A pooled literature; reviews or meta-analyses exist.

RMSSD/pNN50 are among the most heavily studied HRV measures in psychiatry: two independent meta-analyses converge on significantly reduced RMSSD/pNN50 in major depression vs. healthy controls (g ≈ −0.46 to −0.51, thousands of patients pooled), extending to broader cardiovascular-risk and mental-disorder populations.

Dominant reported direction: down: reduced in depression/anxiety (g ≈ −0.3 to −0.5).

Key references: (Koch et al., 2019); (Wu et al., 2023).

Sports & peak performance

Coverage: statistics. A pooled literature; reviews or meta-analyses exist.

RMSSD is the dominant, most-meta-analyzed short-term vagal index in sports science for training-load, recovery and overreaching monitoring, but its direction under overload is genuinely context-dependent: it can rise or fall depending on overload type/timing, an ambiguity documented across multiple competing meta-analyses rather than settled.

Dominant reported direction: context-dependent: rises with some overreaching patterns, falls with others (pre-competition taper).

Key references: (Bellenger et al., 2016).

Contemplative practice

Coverage: individual papers. A small, scattered literature with no pooled meta-analysis.

The one dedicated meta-analysis pooled only 3 studies and found a null RMSSD effect, explicitly citing too few large RCTs, yet several individually-reported trials claim significant RMSSD/pNN50 increases with practice, the largest reported-vs-pooled gap of the three application domains.

Dominant reported direction: contested: meta-analytic null vs. positive individual trials.

Key references: (Rådmark et al., 2019).

See the effect-distribution meta-analysis page for the harvested per-study effect sizes/p-values behind these domain summaries (docs/zoo_gen/effect_stats.csv).

Resources

Resource-intensity rank ◍◍◌◌◌ low is measured: median wall-clock time and allocations over a 360-beat window on synthetic realistic RR (docs/zoo_gen/bench_resources.jl; full grid in resource_bench.csv).

Cold = fresh memoization caches (builds every shared representation from scratch); warm = shared representations (diff, periodogram, Poincaré coords, DFA fluctuation) already cached, so only this feature is recomputed. The tier is derived from the cold cost; see Level 1–2 (base statistics over successive differences).

Citation

Mietus et al. (2002), "The pNNx files": generalised the pNNx family including pNN20.

Seminal reference(s): (Mietus et al., 2002).

See the References page for the full bibliography.