Sleep & CPAP Report — 2026-09-15

🚩 Flags

• Mask-seal spike from 09-14 did NOT repeat — leak fully recovered. Leak_90 = 2.4 L/min (yesterday 4.8, back to 30d avg of 2.2). Leak_Max = 14.4 (yesterday 27.6). Leak_Avg = 0.74 (yesterday 0.86). This confirms 09-14's leak deterioration was an isolated single-night event rather than a persistent mask-fit degradation — the leading suspect from yesterday's report is now cleared for at least one more night. Continue to monitor.
• Flow-limitation load spiked — the most elevated recent-window reading, and it's NOT driven by leak. FL_P95 = 0.12 (highest in the recent 10-night window; prior recent peak was 0.11 on 09-05, 09-07, 09-09, 09-13). Min_Above_010 = 21.2 min (+45.9% vs 7d avg 14.53, +16.3% vs 30d avg 18.23) — the highest since 09-07 (21.3). Min_Above_020 = 6.9 min (+73.1% vs 7d avg 3.99, +62.7% vs 30d avg 4.24) — the highest since 09-13 (7.6). Frac_NonZero (share of samples with any flow-limit signal) = 26.36% — notably above baseline. Leak was fine tonight, so this is upper-airway resistance loading, not mask-seal-driven artifact. Combined with the second-consecutive high P90 (10.02) and highest recent Pressure_Avg (7.76), the machine is being asked to work harder against airway narrowing rather than mask leak.
• AHI 1.93 — first below-baseline night in 4 nights. 7d avg 2.18 (-11.4%), 30d avg 2.01 (-4.2%). Breaks a 3-of-4-night run above the 30d baseline noted in yesterday's report. OA count 10 (near 7d avg 10.43); H count 2 (near 7d avg 2.0); CA count 0 and RERA count 0 — both back to clean baseline after yesterday's flagged elevations (CA 2, RERA 2). Overall event profile is much cleaner than 09-14, except for the flow-limit spike above.
• Pressure_Avg 7.76 — highest in the recent 10-night window (next highest was 09-13's 7.78). Consistent with the machine responding to elevated flow-limitation load rather than to leak (which was low).
• Sleep architecture came in solid. Deep% 12.66% (+9.2% vs 7d avg 11.60, -3.1% vs 30d avg 13.06). REM% 24.62% (+11.2% vs 7d avg 22.15, +4.7% vs 30d avg 23.51). Awake time 8.53 min (-1.9% vs 7d avg 8.70, -5.4% vs 30d avg 9.02). Sleep efficiency 97.69%. The flow-limit elevation did not translate into sleep-stage disruption or fragmentation — consistent with the historical near-zero correlation between Min_Above_010 and deep% (r = +0.020, n=80).
• Time in bed remains short. TIB 369.3 min (6h 9.3m); Total sleep 360.8 min (6h 0.8m). -3.9% vs 30d TIB, but roughly on trend for the recent 7d (7d avg 366.1). CPAP usage 6.23h — closely matches TIB, no early mask removal.

Two-night pattern update: 09-14 flagged leak (CPAP hardware) with intact sleep architecture. 09-15 clears the leak concern but surfaces a different CPAP-side signal — flow-limitation load, which is airway-physiology rather than hardware. Sleep quality remains protected in both. If flow-limitation stays this elevated for 2–3 more nights, that would be a signal worth watching (potentially: allergy/congestion, positional airway, or gradual airway sensitivity shift). One night alone is not yet a trend.

Summary Table — Night of 2026-09-15

Metric	Value
Sleep start	2026-09-15 23:45:40-05:00
Time in bed	369.3 min (6h 9.3m)
Total sleep	360.8 min (6h 0.8m)
Sleep efficiency	97.69%
Awake time	8.53 min
Core (N1+N2) sleep	226.3 min / 62.73%
Deep (N3) sleep	45.66 min / 12.66%
REM	88.81 min / 24.62%
CPAP usage hours	6.23
Session count	1
AHI	1.93
RDI	1.93
Obstructive apneas (OA)	10
Hypopneas (H)	2
Central apneas (CA)	0
Unclassified apneas (UA)	0
RERA	0
Pressure_Avg	7.76
Pressure_Min	4.00
Pressure_Max	12.18
Pressure_90th	10.02
Leak_Avg	0.74
Leak_Max	14.4
Leak_90th	2.4
FL_Avg	0.0186
FL_P95	0.12
FL_Max	0.46
Min_Above_005	43.0
Min_Above_010	21.2
Min_Above_020	6.9
Frac_NonZero (FL samples)	26.36%

Comparison to Trailing Averages

Metric	Tonight	7d avg	30d avg	Δ vs 7d	Δ vs 30d
AHI	1.93	2.18	2.01	-11.4%	-4.2%
OA count	10.00	10.43	9.70	-4.1%	+3.1%
H count	2.00	2.00	2.57	0.0%	-22.1%
CA count	0.00	1.14	0.87	-100.0%	-100.0%
RERA count	0.00	0.71	0.87	-100.0%	-100.0%
Pressure_Avg	7.76	7.31	7.46	+6.2%	+4.0%
Pressure_90	10.02	9.86	10.04	+1.6%	-0.2%
Leak_Avg	0.74	0.61	0.81	+20.7%	-8.3%
Leak_Max	14.40	17.49	17.84	-17.7%	-19.3%
Leak_90th	2.40	1.54	2.20	+55.6%	+9.1%
Usage hours	6.23	6.26	6.51	-0.5%	-4.3%
FL_P95	0.12	0.087	0.099	+37.7%	+21.2%
Min_Above_010	21.20	14.53	18.23	+45.9%	+16.3%
Min_Above_020	6.90	3.99	4.24	+73.1%	+62.7%
Awake min	8.53	8.70	9.02	-1.9%	-5.4%
TIB min	369.31	366.13	384.08	+0.9%	-3.9%
Total sleep min	360.78	357.43	375.07	+0.9%	-3.8%
Deep %	12.66	11.60	13.06	+9.2%	-3.1%
REM %	24.62	22.15	23.51	+11.2%	+4.7%
Core %	62.73	66.26	63.42	-5.3%	-1.1%

Week-over-Week Deltas (recent 7d 09/09–09/15 vs prior 7d 09/02–09/08)

Metric	Recent 7d	Prior 7d	Δ	Δ %
AHI	2.131	2.094	+0.037	+1.8%
OA	10.571	9.714	+0.857	+8.8%
CA	1.143	0.429	+0.714	+166.7%
Pressure_Avg	7.361	7.369	-0.007	-0.1%
Leak_Avg	0.559	0.936	-0.377	-40.3%
Leak_90	1.371	2.743	-1.371	-50.0%
Deep % (Apple)	12.40	12.80	-0.40	-3.1%
REM % (Apple)	23.05	21.73	+1.32	+6.1%
Awake min (Apple)	8.28	10.75	-2.47	-23.0%
Sleep min (Apple)	357.93	369.67	-11.73	-3.2%
FL_P95	0.091	0.093	-0.002	-1.5%
Min_Above_010	15.41	15.91	-0.50	-3.1%
Min_Above_020	4.44	3.46	+0.99	+28.5%

Recent-week leak averages continue to track well BELOW the prior week (Leak_90 -50%, Leak_Max -21%) — the 09-14 spike was the outlier, not a trend inflection. CA count remains elevated week-over-week (+167%), driven largely by 09-10's 3-CA outlier and 09-14's 2 CAs. Min_Above_020 rose +28.5% week-over-week, and tonight's 6.9 pushes that trend further — worth watching. Sleep architecture week-over-week is stable (Deep -3.1%, REM +6.1%, Awake -23%).

Recent 10-Night Context

Date	AHI	OA	CA	PAvg	P90	LAvg	LMax	L90	FL_P95	MinAbv010	MinAbv020	Deep%	REM%	Awake
2026-09-06	1.40	7	1	6.72	9.44	0.27	14.4	0.0	0.08	14.4	3.8	17.64	19.40	10.04
2026-09-07	1.85	9	1	7.23	10.40	1.21	21.6	3.6	0.11	21.3	2.2	8.05	22.47	13.05
2026-09-08	2.25	9	0	7.38	9.34	1.12	18.0	3.6	0.09	15.0	3.7	7.85	19.22	11.04
2026-09-09	1.40	7	1	7.68	10.52	0.63	13.2	1.2	0.11	20.5	4.8	15.10	29.01	2.51
2026-09-10	2.10	10	3	7.14	9.30	0.14	7.2	0.0	0.07	9.7	1.4	11.44	20.34	7.02
2026-09-11	2.53	13	0	6.57	9.52	0.62	24.0	0.0	0.05	8.9	2.4	11.97	21.87	10.54
2026-09-12	1.81	8	1	7.17	9.92	0.45	14.4	1.2	0.08	12.7	3.9	18.06	27.63	9.53
2026-09-13	2.60	14	1	7.78	10.16	0.47	18.0	0.0	0.11	19.1	7.6	8.29	20.78	18.06
2026-09-14	2.55	12	2	7.43	10.26	0.86	27.6	4.8	0.10	15.8	4.1	14.93	21.67	3.01
2026-09-15	1.93	10	0	7.76	10.02	0.74	14.4	2.4	0.12	21.2	6.9	12.66	24.62	8.53

CPAP → Sleep-Stage Correlations (all 83 matched nights)

Correlations computed across full available history (2026-06-17 → 2026-09-15).

Pair	the user r	Interpretation
AHI vs deep_pct	+0.023	negligible
AHI vs awake_min	+0.059	negligible
OA vs awake_min	+0.093	negligible
CA vs deep_pct	+0.033	negligible
Leak_Avg vs deep_pct	-0.154	weak negative — higher leak, slightly less deep sleep
Leak_90 vs deep_pct	-0.178	weak negative — same direction
Leak_Max vs awake_min	+0.062	negligible
FL_P95 vs awake_min	-0.080	negligible
Min_Above_010 vs deep_pct	+0.020	negligible

Bucket comparison (Leak_90 median split, median = 2.4): - Low-leak nights (n=49): mean deep% 15.33, awake 15.25 min - High-leak nights (n=24): mean deep% 12.60, awake 13.09 min - Δ (high – low): deep% -2.73 pp, awake -2.16 min

Tonight (2026-09-15) sits right at the leak median (Leak_90 = 2.4) and had a good sleep-stage outcome: Deep% 12.66 is near the high-leak-bucket mean, and Awake 8.53 min is BELOW both buckets' averages. The much larger deviation tonight is on flow-limitation, and correlations show Min_Above_010 has near-zero predictive power for deep% (r = +0.020), so the elevated FL should NOT be expected to have degraded sleep quality — and indeed did not. The correlation table continues to say leak is the CPAP-side variable most weakly linked to sleep architecture; flow limitation is even weaker on average.

Caveats

• Sleep session started at 23:45:40 on 2026-09-15 — matched to the 2026-09-15 CPAP file (same-calendar-date convention).
• Apple Watch SpO2/pulse channels not present in current CPAP CSV export.
• Correlations are directional/small-effect only — single-night deviations often exceed what correlations alone would predict.
• All flagged metrics remain within clinically safe ranges (AHI < 5, leak within mask spec); the flags mark deviations from the user's own recent baseline, not clinical thresholds.
• The health-report-2026-07-18-to-2026-07-19 2.json duplicate file was again ignored in favor of the primary copy.
• Matched-night count is 83 (up from 75 in prior report). Backfill from earlier 2026-06 nights (via the multi-day 2026-06-15_2026-07-10.csv roll-up) is present but those early nights have zero-valued pressure and leak channels; these should still be treated as low-confidence for CPAP metrics prior to ~2026-06-22.