Therapy Pressure, Leak
& Residual Events

Hours of delivered therapy for the night — the denominator behind every per-hour event index. Mask-on is delivered pressure > 0; brief gaps under ~5 min are not counted as off.

Percentage of nights in the trailing 30 days with at least 4 hours of use — the adherence figure insurers and clinicians track.

Minutes from recording start to the first stable delivered pressure — how long the mask sat unused before therapy began.

The median delivered pressure across mask-on time — the typical pressure the airway needed.

The pressure exceeded only 5% of the night — the standard summary of an APAP machine’s pressure demand.

Interquartile spread of delivered pressure — how much an auto-titrating machine moved through the night.

Median drop from inspiratory to expiratory pressure — how much the machine eases pressure on exhale for comfort.

P95 of the expiratory pressure — from a dedicated EPAP lane on bilevel machines, or derived from EPR-relieved expiratory pressure on CPAP/APAP.

Obstructive + central apneas + hypopneas per hour of use — the headline of how well therapy is controlling sleep-disordered breathing tonight.

Central apneas per hour of use. Elevated central events on CPAP can signal treatment-emergent central apnea (TECA) — which may call for a different therapy mode (e.g. ASV) rather than more pressure.

Obstructive apneas per hour. Persisting obstructive events suggest the delivered pressure may be inadequate.

Partial-airflow reductions per hour. Hypopneas dominate most residual AHI.

Arousals from increased respiratory effort that fall short of an apnea/hypopnea, per hour. Estimated from flow-limitation shape; not all machines report it.

Share of therapy time scored as Cheyne–Stokes respiration or periodic breathing (waxing–waning crescendo–decrescendo airflow), from the CSL annotation stream.

Mean of the machine’s flow-limitation index — a continuous 0–1 measure of inspiratory-flow flattening (a precursor to obstruction). Device-derived and less standardized across manufacturers.

Fraction of therapy time the flow-limitation index exceeded 0.3 — how much of the night the airway was meaningfully narrowed without yet triggering a scored event.

Share of therapy time the device’s snore index exceeded 0.2 — airway vibration on the pressure/flow signal. Persistent snore can indicate sub-optimal pressure.

Pearson correlation between snore index and delivered pressure — a CPAPDex composite probing whether more pressure is suppressing snore (negative) or failing to (flat/positive). Directional, internal.

Typical excess mask leak across the night. Low and steady is the goal.

Peak leak exceeded only 5% of the night — captures intermittent seal breaks a median can hide.

The single highest leak value of the night — worst-case seal break, useful context for brief data-quality dropouts.

Fraction of the night above ResMed’s 24 L/min large-leak threshold — the trust gauge for everything else. A high value means the residual AHI should be read with heavy skepticism.

Standard deviation of leak divided by its mean — a CPAPDex composite for seal stability rather than level. A steady moderate leak scores better than an erratic one.

Median breaths per minute reported by the machine. Normal adult sleeping rate is ~12–18.

Interquartile spread of breathing rate — a steadiness measure. A wide range can reflect periodic breathing or fragmented sleep.

Median estimated volume per breath. The machine estimates this from flow — a relative trend indicator, not calibrated spirometry.

Median air moved per minute (tidal volume × respiratory rate) — useful for spotting hypoventilation trends across nights.

Coefficient of variation of minute ventilation — a CPAPDex composite flagging unstable breathing (e.g. periodic-breathing crescendos) even when the median looks normal.

Total breaths CPAPDex detected from the raw flow waveform — the basis for the flow-derived rate and a data-completeness check.

Breaths per minute from CPAPDex’s own flow detection — an independent cross-check against the machine’s reported respiratory rate. Close agreement validates both.

Ratio of inspiratory to expiratory flow time per breath. Shifts can reflect airway resistance or expiratory effort; method-dependent.

Self-gated AASM 3% desaturation events per hour — an independent corroboration of residual events from the oxygen side.

Share of valid oximetry samples below 90% saturation — the standard cumulative-hypoxemia burden. Near zero on effective therapy.

The lowest valid SpO₂ reading after self-gating — the worst oxygen point that survived artifact rejection.

Average valid oxygen saturation across the night — the baseline oxygenation therapy is sustaining.

Median heart rate from the oximeter (physiologically plausible in-band values) — context for the oxygen trace, not a substitute for the cardiac nodes.

Spread of the in-band pulse rate — a coarse variability indicator; wide swings can accompany arousals or residual events.

CPAPDex decodes the EDF header (labels, sample rates, physical scaling) and each signal’s records, then parses the EVE/CSL annotation TALs into classed events with absolute timestamps. Physical units are validated on decode.

The EDF start date/time maps to a floating wall-clock anchor per the Clock Contract (read back with UTC getters so displayed time is viewer-timezone-independent). A night can contain multiple sessions separated by an off-mask gap — CPAPDex stitches them while preserving the gap.

On the SA2 lane, CPAPDex applies the same self-gate as OxyDex before counting desaturations — rejecting perfusion-collapse and edge artifacts so a squeeze-artifact dip never inflates the ODI. Missing samples surface as gaps, never fabricated.