Sleep & CPAP Report — 2026-09-25

🚩 Flags

• Deep sleep 21.66% — top decile, 30d z = +1.94. Vs 30d avg 13.41% (+61.6%) and 7d 12.57% (+72.4%). Absolute deep 73.3 min, well above 30d mean ~50 min. This decisively breaks the week-over-week deep-sleep decline flagged in the 2026-09-24 report — that trend called for watching; tonight is the strongest single-night reversal in the recent series (prior 10-night high was 18.83% on 2026-09-23).
• ⚠️ CPAP usage 5.83 h — shortest session in the trailing 30 days (30d z = -1.62). Total sleep 338.2 min (5h 38m), lowest in the recent 10-night window. 8 consecutive nights of declining CPAP hours (6.87 → 6.18 → 6.42 → 6.20 → 5.93 → 6.18 → 6.05 → 5.83). Not clinically short, but the sustained downward trend is notable — worth checking whether bedtime or wake-time has been drifting this week.
• ⚠️ Pressure elevated: P_avg 8.04, P_90 10.60 — highest in recent memory (30d z = +1.28 / +1.30). Coincides with 11 obstructive apneas (vs 7d 8.0) and slightly higher FL_P95 (0.11 vs 30d 0.09). The pressure algorithm is working harder tonight — consistent with a body-position or airway-loading issue rather than a mask problem.
• AHI 2.74 — moderately elevated but well under threshold. Vs 7d 2.14 (+27.8%) and 30d 2.18 (+25.6%); 30d z = +0.82. Driven by 11 OA (matches 30d avg 10.1) + 4 H + 1 CA + 0 RERA. No individual event class is anomalous — this is a "sum of small parts" AHI, not a spike from any single event category.
• Leak elevated week-over-week: Leak_avg 1.13, Leak_90 3.60. Tonight fits the pattern of the last 7 days (7d avg leak_avg 1.12, leak_90 3.43) but stays +38% and +50% above 30d baseline. Recent week vs prior week: Leak_avg +165%, Leak_90 +188% — this is now a sustained shift, not a one-night blip. No sign of a runaway leak session tonight (Leak_Max 22.80 is in the middle of the recent range).
• Sleep efficiency 99.12%; awake only 3.0 min. Second-lowest awake time in the recent 10 nights (only 2026-09-24 at 1.0 min was lower). Continuity was excellent given the short total time in bed.

Summary Table — Night of 2026-09-25

Metric	Value
Sleep start	2026-09-26T00:05:36-05:00
Time in bed	341.22 min (5h 41.2m)
Total sleep	338.21 min (5h 38.2m)
Sleep efficiency	99.12%
Awake time	3.01 min
Core (N1+N2) sleep	184.16 min / 54.45%
Deep (N3) sleep	73.26 min / 21.66%
REM	80.79 min / 23.89%
CPAP usage hours	5.83
Session count	1
AHI	2.74
RDI	2.74
Obstructive apneas (OA)	11
Hypopneas (H)	4
Central apneas (CA)	1
Unclassified apneas (UA)	0
RERA	0
Pressure_Avg	8.04
Pressure_Min	4.00
Pressure_Max	12.30
Pressure_90th	10.60
Leak_Avg	1.13
Leak_Max	22.80
Leak_90th	3.60
FL_Avg	0.0165
FL_P95	0.11
FL_Max	0.60
Min_Above_005	38.5
Min_Above_010	18.0
Min_Above_020	3.1
Frac_NonZero (FL samples)	25.06%

Comparison to Trailing Averages

Metric	Tonight	7d avg	30d avg	Δ vs 7d	Δ vs 30d
AHI	2.74	2.14	2.18	+27.8%	+25.6%
OA count	11.00	8.00	10.10	+37.5%	+8.9%
H count	4.00	5.00	2.97	-20.0%	+34.8%
CA count	1.00	0.29	0.97	+250.0%	+3.4%
RERA count	0.00	0.57	0.73	-100.0%	-100.0%
Pressure_Avg	8.04	7.62	7.45	+5.5%	+8.0%
Pressure_90	10.60	10.11	9.95	+4.9%	+6.5%
Leak_Avg	1.13	1.12	0.82	+1.0%	+38.1%
Leak_Max	22.80	22.29	18.44	+2.3%	+23.6%
Leak_90th	3.60	3.43	2.40	+5.0%	+50.0%
Usage hours	5.83	6.26	6.44	-6.9%	-9.5%
FL_P95	0.11	0.10	0.09	+10.0%	+18.3%
Min_Above_010	18.00	17.91	16.44	+0.5%	+9.5%
Min_Above_020	3.10	3.53	3.71	-12.1%	-16.4%
Awake min	3.01	7.96	8.98	-62.2%	-66.5%
TIB min	341.22	373.76	380.30	-8.7%	-10.3%
Total sleep min	338.21	365.80	371.32	-7.5%	-8.9%
Deep %	21.66	12.57	13.41	+72.4%	+61.6%
REM %	23.89	25.32	24.12	-5.7%	-1.0%
Core %	54.45	62.11	62.47	-12.3%	-12.8%

Week-over-Week Deltas (recent 7d 2026-09-19–2026-09-25 vs prior 7d 2026-09-12–2026-09-18)

Metric	Recent 7d	Prior 7d	Δ	Δ %
AHI	2.411	2.146	+0.266	+12.4%
OA	8.857	10.857	-2.000	-18.4%
CA	0.429	1.429	-1.000	-70.0%
Pressure_Avg	7.757	7.369	+0.389	+5.3%
Leak_Avg	1.254	0.473	+0.781	+165.3%
Leak_90	3.943	1.371	+2.571	+187.5%
Deep % (Apple)	14.018	13.351	+0.667	+5.0%
REM % (Apple)	25.422	23.031	+2.391	+10.4%
Awake min (Apple)	6.882	11.398	-4.516	-39.6%
Sleep min (Apple)	358.703	368.097	-9.394	-2.6%
FL_P95	0.103	0.093	+0.010	+10.8%
Min_Above_010	18.143	15.814	+2.329	+14.7%
Min_Above_020	3.557	4.500	-0.943	-21.0%

Recent 10-Night Context

Date	AHI	OA	CA	PAvg	P90	LAvg	LMax	L90	FL_P95	MinAbv010	MinAbv020	Deep%	REM%	Awake
2026-09-16	2.85	16	4	7.20	9.68	0.09	8.40	0.00	0.08	13.5	2.8	9.37	22.95	10.04
2026-09-17	2.41	11	2	7.17	9.22	0.52	10.80	1.20	0.07	12.0	3.3	18.64	20.41	20.07
2026-09-18	0.87	5	0	7.07	9.56	0.18	8.40	0.00	0.09	16.4	2.9	11.51	23.16	10.54
2026-09-19	3.23	15	0	7.60	9.64	0.41	14.40	1.20	0.10	16.8	2.1	10.83	25.42	14.55
2026-09-20	2.18	7	0	8.21	10.54	1.85	27.60	6.00	0.13	28.6	7.0	8.70	31.40	14.55
2026-09-21	1.94	5	1	7.82	10.46	0.90	14.40	1.20	0.09	16.2	2.9	13.07	23.78	5.52
2026-09-22	3.20	12	0	7.49	10.32	1.03	36.00	2.40	0.10	14.8	3.1	15.33	23.07	4.01
2026-09-23	1.94	9	0	7.78	10.06	2.87	26.40	12.00	0.10	17.4	3.3	18.83	26.11	5.52
2026-09-24	1.65	3	1	7.36	10.18	0.59	28.80	1.20	0.09	15.2	3.4	9.69	24.30	1.00
2026-09-25	2.74	11	1	8.04	10.60	1.13	22.80	3.60	0.11	18.0	3.1	21.66	23.89	3.01

CPAP → Sleep-Stage Correlations (all 71 matched nights)

Correlations across the full available history (2026-07-13 → 2026-09-25) where both a CPAP daily row and a per-night Apple Watch sleep-stage session exist.

Pair	the user r	Interpretation
AHI vs deep_pct	-0.113	weak negative
AHI vs awake_min	-0.183	weak negative
OA vs awake_min	-0.054	negligible
CA vs deep_pct	-0.049	negligible
Leak_Avg vs deep_pct	-0.005	negligible
Leak_90th vs deep_pct	-0.010	negligible
Leak_Max vs awake_min	-0.201	weak negative
FL_P95 vs awake_min	-0.063	negligible
Min_Above_010 vs deep_pct	-0.180	weak negative

Bucket comparison (Leak_90 median split, median = 1.20): - Low-leak nights (n=37): mean deep% 14.25, awake 14.42 min - High-leak nights (n=34): mean deep% 13.97, awake 11.82 min - Δ (high – low): deep% -0.28 pp, awake -2.60 min

Tonight sits in the high-leak bucket (L90 3.60) yet delivered the best deep-sleep percentage of the last 10 nights — a direct counter-example to any strong "leak drives shallow sleep" hypothesis. Correlations remain weak across the board; single-night variation dominates.

Interpretation

Tonight is a short but architecturally excellent night. The last time deep-sleep percentage was this high in the tracked history is not visible in the recent 10-night context (peak there was 18.83% on 2026-09-23), and the z-score of +1.94 vs 30d puts it in roughly the top 3–5% of nights. Absolute deep-sleep minutes (73.3) are the highest since the deep-sleep decline began. This is a strong, unambiguous reversal of the week-long deep-sleep dip flagged yesterday.

Three things pull in the other direction:

1. Duration is the real story. 5.83 CPAP hours is the shortest in 30 days, and it caps an eight-night steady decline in CPAP usage. Total sleep of 5h 38m is below what most adults need for sustained cognitive function, especially when repeated. The night's high deep-sleep share partly reflects the shortened window (deep sleep concentrates in the first cycles, so a truncated night can over-index on it), but the absolute deep minutes are genuinely high — so this is not purely a percentage artifact. The concern is not any single night; it is that the trailing 8-night run of shorter and shorter sessions may be shifting the overall sleep-need trajectory.

2. Pressure and event mix suggest the airway loaded up tonight. P_avg 8.04 and P_90 10.60 are the highest in the visible window, and 11 OA is the highest OA count in the last 6 nights. Neither is clinically concerning in isolation, but the pattern — higher pressure, more obstructive events, slightly higher FL — is consistent with either a supine-heavy sleep posture or an evening trigger (alcohol, later-than-usual meal, congestion). The auto-titration is doing its job; nothing to intervene on.

3. Leak has re-established a higher weekly baseline. Week-over-week leak metrics are up 165%–188%. Tonight itself was not a leak spike (Leak_Max 22.8 is unremarkable), but the seven-day pattern means the mask fit or straps may deserve a fresh check. Worth watching whether this week's leak elevation persists into next week or drifts back toward the 2026-09-12–18 baseline.

Overall: a short, dense, efficient, deep-rich night. Zero clinical red flags. The one metric to keep watching over the next few days is total sleep duration.

Caveats

• Sleep session started at 2026-09-26T00:05:36-05:00 — matched to the 2026-09-25 CPAP file, consistent with the established convention (CPAP Date = calendar date of the sleep session's start day, aligned to timeWindow.start).
• 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, and tonight explicitly bucks the (weak) leak/deep-sleep pattern.
• All flagged metrics remain within clinically safe ranges (AHI < 5, leak within mask spec, usage 5+ hours); flags mark deviations from the user's own recent baseline, not clinical thresholds.
• Matched-night count this report = 71, unchanged from 2026-09-24 (the pipeline continues to pair one CPAP row per calendar date with one Apple Watch session where both exist).
• The high deep-sleep share benefits from a shorter TIB window (deep concentrates early in the night), but absolute deep minutes are also elevated — the improvement is real, not a percentage-only artifact.
• 30d and 7d averages here exclude tonight's night by construction (they cover the 30 and 7 nights ending 2026-09-24).