Sleep & CPAP Report — 2026-09-11

🚩 Flags

• Deep-sleep WoW regime shift deepens — recent 7d (09/05–09/11) 11.06% vs prior 7d (08/29–09/04) 16.04% = -5.0 pp (-31.0%). Yesterday's shift was -4.4 pp; tonight it widens to -5.0 pp. Second consecutive report where this is the highest-priority signal. This is no longer a "recent softness" — it's a two-week regime change: absolute deep-sleep minutes have dropped roughly 15 min/night from where they were 10–14 days ago. Not a CPAP mechanic issue; the therapy numbers are actually improving in parallel.
• OA count 13 — HIGHEST of recent 10-night window. Prior 10-night range was 5–11 (median 9). Tonight is +30% above the 7d avg (9.86) and +35% above 30d (9.60). Coupled with the two-week high AHI of 2.53, but note AHI itself is still well within clinically fine (<5). The OA count spike is worth flagging even so — the airway did more obstructive work tonight than any other night in the recent window.
• Flow-limit exposure hit a NEW recent-window low — Min_Above_010 = 8.9 min, FL_P95 = 0.05, both lowest of the recent 10-night window (prior low was yesterday at 9.7 min / 0.07). Min_Above_010 -44% vs 7d (16.01), -52% vs 30d (18.36). Second consecutive night at recent-window lows — this is now durable enough to fully de-escalate the FL flag that dominated the last week of reports.
• Pressure at a recent-window low — Pressure_Avg 6.57 (LOWEST of recent 10), Pressure_90 9.52 (LOWEST of recent 10). Pressure_Avg -10% vs 7d (7.30), -12% vs 30d (7.50). The AutoSet continues to relax in response to the improved FL/leak picture. Pressure_Max 10.86 is also the lowest max of the recent 10 (prior 10-night range: 11.87–13.53). Genuinely the calmest pressure night of the window.
• Zero central apneas — first zero-CA night in 6 nights. Last 6 nights had CA counts of 1, 3, 1, 0, 1, 1, 1 → tonight breaks the streak. Note: the WoW CA number still reads +133% (recent 7d 1.00 vs prior 7d 0.429), because tonight's zero doesn't reverse a full week of 1's — but the tonight-vs-yesterday comparison (0 vs 3) is a clean directional improvement.
• Leak_Max 24 L/min — highest single-sample leak of the recent 10-night window (prior 10-night max range: 4.8–17.6). However Leak_Avg 0.62 and Leak_90th 0.00 both remain in the well-controlled band. This looks like one brief mask-position event that didn't affect the overall session — but the peak is worth noting after two nights of essentially zero leak activity.
• Leak seal remains durable at the aggregate level — Leak_90th 0.00 for the 3rd consecutive night. Recent 7d Leak_Avg 0.68 vs prior 7d 1.00 = -31.5% WoW (essentially unchanged from yesterday's -35.7%). The Leak flag stays fully de-escalated.

Summary Table — Night of 2026-09-11

Metric	Value
Sleep start	2026-09-11 23:49:17-05:00
Time in bed	375.3 min (6h 15.3m)
Total sleep	364.8 min (6h 4.8m)
Sleep efficiency	97.19%
Awake time	10.5 min
Core (N1+N2) sleep	241.4 min / 66.16%
Deep (N3) sleep	43.7 min / 11.97%
REM	79.8 min / 21.87%
CPAP usage hours	6.32
Session count	1
AHI	2.53
RDI	2.53
Obstructive apneas (OA)	13
Hypopneas (H)	3
Central apneas (CA)	0
Unclassified apneas (UA)	0
RERA	0
Pressure avg	6.57 cmH₂O
Pressure min	4.00 cmH₂O
Pressure 90th	9.52 cmH₂O
Pressure max	10.86 cmH₂O
Leak avg	0.62 L/min
Leak 90th	0.0 L/min
Leak max	24.0 L/min
Flow-limit p95	0.05
Flow-limit p99	0.17
Flow-limit max	0.34
Flow-limit min >0.10	8.9
Flow-limit min >0.20	2.4

Trailing Averages Comparison

Metric	Tonight	7-day avg	Δ vs 7d	30-day avg	Δ vs 30d
AHI	2.53	2.15	+18%	1.97	+28%
OA count	13	9.86	+32%	9.60	+35%
H count	3	2.71	+11%	2.57	+17%
CA count	0	1.00	-100%	0.87	-100%
Leak_Avg (L/min)	0.62	0.64	-3%	0.76	-18%
Leak_90th (L/min)	0.0	1.54	-100%	2.12	-100%
Pressure_Avg	6.57	7.30	-10%	7.50	-12%
Pressure_90th	9.52	9.93	-4%	10.06	-5%
Usage hours	6.32	6.34	0%	6.62	-5%
FL min >0.10	8.9	16.01	-44%	18.36	-52%
FL min >0.20	2.4	3.67	-35%	4.05	-41%
FL_P95	0.05	0.09	-44%	0.10	-50%
Deep sleep %	11.97	11.44	+0.5 pp	13.29	-1.3 pp
REM %	21.87	22.60	-0.7 pp	23.93	-2.1 pp
Awake min	10.5	9.10	+16%	11.09	-5%
Total sleep (min)	365	367	-1%	380	-4%

Week-over-Week Trend

Metric	Recent 7d (09/05–09/11)	Prior 7d (08/29–09/04)	Direction
AHI	2.03	2.17	↓ -6.5%
OA count	9.43	11.00	↓ -14.3%
H count	2.43	2.86	↓ -15.0%
CA count	1.00	0.43	↑ +133.3%
Leak_Avg	0.68	1.00	↓ -31.5%
Leak_90	1.54	3.43	↓ -55.0%
Pressure_Avg	7.19	7.78	↓ -7.6%
Pressure_90	9.83	10.23	→ -4.0%
FL min >0.10	15.66	18.89	↓ -17.1%
Deep sleep %	11.06	16.04	↓ -31.0% (-5.0 pp)
REM %	22.52	23.84	↓ -5.5%
Total sleep (min)	367.0	379.8	↓ -3.4%

The WoW picture keeps splitting cleanly along the same axis: CPAP mechanics improving (AHI ↓, OA ↓, H ↓, Leak ↓, Pressure ↓, FL ↓ all vs prior 7d), sleep architecture worsening (Deep -5.0 pp, REM -5.5%). The CA WoW spike (+133%) looks alarming but is amplified by a low prior-window base (0.43 CA/night); absolute rates remain well below clinically concerning thresholds. The deep-sleep WoW gap is the durable, quantitatively significant signal.

Recent 10-Night Detail

Date	AHI	OA	H	CA	Leak_Avg	Leak_90	P_Avg	Hours	Deep%	Awake min	FL_p95	Min>0.10
2026-09-11	2.53	13	3	0	0.62	0.0	6.57	6.32	11.97	10.5	0.05	8.9
2026-09-10	2.10	10	0	3	0.14	0.0	7.14	6.20	11.44	7.0	0.07	9.7
2026-09-09	1.40	7	1	1	0.63	1.2	7.68	6.42	15.10	2.5	0.11	20.5
2026-09-08	2.25	9	5	0	1.12	3.6	7.38	6.23	7.85	11.0	0.09	15.0
2026-09-07	1.85	9	2	1	1.21	3.6	7.23	6.48	8.05	13.0	0.11	21.3
2026-09-06	1.40	7	1	1	0.27	0.0	6.72	6.45	17.64	10.0	0.08	14.4
2026-09-05	2.66	11	5	1	0.79	2.4	7.59	6.38	5.35	13.5	0.11	19.8
2026-09-04	3.37	16	5	0	0.31	0.0	7.37	6.23	14.66	6.5	0.07	11.4
2026-09-03	2.15	11	4	0	1.98	7.2	7.65	6.97	15.66	15.6	0.10	17.1
2026-09-02	0.98	5	1	0	0.87	2.4	7.64	6.12	20.37	5.5	0.09	12.4

Correlation Analysis (68 matched nights)

Pair	the user r
AHI ↔ Deep sleep %	+0.09 (none)
AHI ↔ Awake minutes	+0.06 (none)
Leak_Avg ↔ Deep sleep %	-0.18 (weak negative)
Leak_Avg ↔ Awake minutes	-0.02 (none)
Leak_90th ↔ Deep sleep %	-0.21 (weak negative)
Leak_90th ↔ Awake minutes	-0.02 (none)

High- vs low-leak bucket comparison (split at Leak_Avg median = 0.77 L/min):

• Low-leak nights (n=35): mean deep sleep 15.22%, awake 15.7 min
• High-leak nights (n=33): mean deep sleep 13.92%, awake 13.8 min

Tonight (Leak_Avg 0.62) is in the low-leak bucket — the group whose mean deep% is 15.22%. Actual tonight was 11.97%, a -3.2 pp shortfall vs the low-leak expectation. Third consecutive night where a low-leak seal has not delivered proportional deep sleep. The Leak↔Deep correlation across the whole dataset (r=-0.18) remains a weak-negative directional signal only — most of the deep-sleep variance is elsewhere.

Interpretation

Another mechanically strong CPAP night, another mediocre architecture night. Same pattern as yesterday, and now clearly the modal pattern of the past week. Tonight the CPAP profile is arguably even better than yesterday's: pressure at a recent low (P_Avg 6.57), FL exposure at a NEW recent-window low (Min_Above_010 8.9), leak flat, zero centrals. Every mechanical improvement flag from yesterday held or strengthened. Yet deep-sleep % is again below the 30d baseline, and REM is again below both.

The two-story split is now stable enough to call it. Recent 7d Deep% 11.06 vs prior 7d 16.04 (-5.0 pp) is a substantial regime change. In absolute terms, the past week has averaged ~40 min/night of deep sleep vs ~60 min/night in the prior week — roughly a third less N3 while therapy compliance and CPAP quality metrics are unchanged or improved. The sleep-architecture story is running on a different clock than the CPAP-mechanics story.

Tonight's one meaningful CPAP-side outlier: OA count 13. This is the highest single-night OA count of the recent 10-window. Because Hours (6.32) and total sleep are near normal, the OA rate per hour is proportionally higher too. Two ways to read this: (a) the AutoSet's decision to run lower pressure tonight allowed more obstructive breakthroughs; or (b) something changed in the airway/position tonight. AHI is still 2.53 — clinically fine — but the OA count deserves noting after a week where the machine and airway have been mostly quieter. If this repeats over the next 2–3 nights alongside continued low pressure, it may be an argument that pressure has eased too much. One night isn't enough to conclude anything.

Leak_Max 24 is an artifact worth naming but not worrying about. Leak_Avg 0.62 and Leak_90 0.00 mean the leak was essentially zero for 90+% of the night, with a single brief peak — likely a momentary mask-position shift (rolling over, adjusting). The event didn't propagate into the aggregate, so seal integrity is fine. Would only matter if Leak_Max jumps like this repeat across nights.

FL flag can now be de-escalated. Two consecutive nights of Min_Above_010 below 10 min, both new recent-window lows, both accompanied by lower pressure and lower AHI-adjacent metrics. The concern that dominated reports for a full week — that the airway was doing more narrowing work than usual — has resolved cleanly for at least this two-night window. If the next 1–2 nights don't spike back up, the flag can be closed entirely.

CA zero-count is a nice reset. Yesterday's 3-CA outlier looks isolated: tonight is 0. The WoW average is still elevated (recent 7d 1.00 vs prior 7d 0.43) because of the last week's cluster, but the tonight-vs-yesterday direction is +good. Continue to watch — if the next 3–5 nights average <0.5 CA, the CA flag can also de-escalate.

Sleep total 6h 4.8m — up 10 min from yesterday, near the 7d avg. Not a duration-constrained night this time. Deep sleep 43.7 min in absolute terms is up 3 min vs yesterday's 40.6 min, but as a fraction of a slightly longer night, deep% barely moved (11.97 vs 11.44). This means the deep-sleep suppression isn't a duration artifact tonight — it's proportional. When the user sleeps a normal length, he's still getting deep-sleep percentages ~2 pp below his 30d baseline.

What the deep-sleep suppression could be (all speculative — the data doesn't distinguish): stress/cortisol elevation; caffeine or alcohol timing shifts; illness or subclinical inflammation; ambient temperature or bedroom conditions; medication changes; something upstream of the airway that CPAP doesn't address. The correlations against every CPAP metric are ≤|0.2| — the mechanical CPAP world has very little explanatory power for tonight's architecture. Whatever's driving this is on the biology side.

Watch items updated: - Deep-sleep suppression — flag STAYS AT HIGHEST PRIORITY (widening). WoW gap grew from -4.4 pp to -5.0 pp. Two-week regime change confirmed. No CPAP-mechanic explanation. Consider non-CPAP factors: sleep timing, alcohol/caffeine, stress, illness, room temperature, meds. - REM % — flag STAYS ELEVATED. 21.87% tonight, below 30d (23.93). Recent 7d REM (22.52) vs prior 7d (23.84) = -5.5% WoW. Smaller effect than deep, same direction. - OA count — new short-term watch item. Tonight's 13 is the recent-window high. Single night — watch next 2–3 nights to see if this pairs with continued low pressure or was a one-off. - FL exposure — flag DE-ESCALATED. Two consecutive recent-window-low nights. Watch one more, then close. - CA count — flag SOFTENED. Tonight zero; yesterday's 3 was an isolated spike. Watch 3–5 more nights. - Leak — flag STAYS DE-ESCALATED. Third night with Leak_90 = 0.0. Leak_Max 24 is a one-second event, not a seal issue. - Pressure — no longer escalating; actually easing. Pressure_Avg dropped from yesterday's 7.14 to 6.57. Pressure trend is now decisively downward. - Short sleep — de-listed. Tonight's 6h 4.8m is close to 7d avg; not a constraint tonight.

Subjectively: If the user woke feeling mediocre or under-rested this morning, the sleep-architecture side supports that — 12% deep and 22% REM is a middling night, though not a bad one. The CPAP side of the ledger is arguably the best it's been all week. This continues to feel like a biology-side issue that the therapy isn't reaching.

Caveats

• SpO2/pulse columns remain empty (device not recording those channels).
• Sleep-stage percentages are Apple Watch estimates (accelerometer + HR), not lab EEG.
• OA=13 is a real count but AHI 2.53 (events/hour) is well within clinically well-controlled range (<5).
• Single CPAP session; matched to Apple Watch by calendar date of sleep-session start timestamp.
• Leak_Max 24 L/min appears to be an instantaneous peak; Leak_Avg 0.62 and Leak_90 0.0 mean the peak did not sustain.
• Personal baselines used for thresholds computed across 68 matched nights: AHI mean+1sd ≈ 3.24; Leak_Avg mean+1sd ≈ 1.98; Deep% mean-1sd ≈ 9.44 (tonight's 11.97% is above that floor but well below the mean of ~14.5%).
• Correlation analysis n=68 matched nights (unchanged vs yesterday — tonight's night was already included in yesterday's dataset via file matching timing; recount holds).