Daily Report — 2026-08-18

Sleep & CPAP Report — Night of 2026-08-18

Sources: health-report-2026-08-18-to-2026-08-19.json (Apple Watch sleep stages, exported 08:23 on 08-19) + OSCAR_Greg_by_Day_2026-08-18.csv (ResMed AirSense/AirFlow 11 via OSCAR). Session start was 23:09 on 08-18 — two minutes later than 08-17's 23:07, holding the third consecutive on-schedule bedtime. CPAP hours 6.12 vs. Watch TIB 6.17 (Δ ≈ 3.1 min), the tightest alignment of the last three nights.

⚠️ Flags

Sleep architecture compressed hard tonight — deep AND REM both fell into the bottom-4 of matched history. Deep 8.86% (rank 4/31 lowest; only 08-01's 5.82%, 07-20's 8.31%, and 08-04's 8.54% were worse); REM 18.71% (rank 3/31 lowest of the 31-night matched sample). Core ballooned to 72.43% — well outside the ~55–65% typical band. Two-night composite deep (08-17 + 08-18): 11.54%, versus the 30-day baseline of 14.56%. The 08-17 report characterized the low-deep readings of 08-13/14 (10.93%, 9.73%) as "a two-night dip rather than a regime shift" and declared architecture "back to baseline." Tonight materially undermines that conclusion — this is now a 3-night pattern with only 08-16/17 as clean outliers in the middle. Combined with tonight's low REM, the interpretation shifts from "occasional single-night dip" to "recurring compression."
AHI cluster did not close on the third night. AHI rebounded 1.11 → 1.63 (+47%); OA rebounded 5 → 7 (+40%); CA (1) and RERA (1) both re-appeared after 08-17's fully clean respiratory profile. The 08-17 report set the disambiguation threshold as "AHI ≤ 1.5 and OA ≤ 6 = cluster closes; AHI ≥ 2.5 or OA ≥ 10 = cluster reopens." Tonight is in between — a partial giveback, not a full relapse. But the 2-night composite AHI ≈ 1.1 improvement is not now a 3-night confirmation; the 08-11 → 08-15 window cannot be closed as a discrete event.
Auto-titrator responded — pressure 90th %ile jumped +1.57 cmH₂O in one night. P_90 climbed from 9.54 → 11.11 cmH₂O, the highest 90th-%ile reading in 8 nights and rank 29/31 (near-top) of the matched history. P_avg is nearly flat (7.57 → 7.55), meaning the machine's baseline demand is unchanged but its tail response is being pushed — it built pressure in response to specific events rather than sustaining a higher baseline. This is the algorithm's own confirmation that the load returned tonight after two calm nights.
Leak now on a 4-night rising streak, 2nd consecutive night ≥ 1.0 — one more night to trigger the mask-fit action item. Nightly leak avg: 0.31 → 0.91 → 1.14 → 1.19 L/min (08-15 → 08-18). Tonight is the 2nd consecutive night ≥1.0; the 08-17 report set the mask-check trigger at 3 consecutive ≥1.0, so 08-19 will decide it. Unlike 08-17 (where the tail dropped to Leak_90 1.2), the tail expanded tonight: Leak_90 jumped from 1.2 → 3.6 L/min (+200%), rank 26/31. This shifts the pattern away from "persistent small-magnitude drift" toward "baseline elevation + return of spikes." If 08-19 leak avg is again ≥1.0 (regardless of the tail), do a strap and cushion check before that night. If 08-19 leak avg drops below 0.5, treat this as a 3–4-night fit adjustment that self-corrected.
Bedtime discipline is now solidly established. Three consecutive nights on schedule (23:02, 23:07, 23:09). The 23:35–23:46 late window of 08-11 → 08-14 is fully behind. Whatever explains tonight's architecture compression, it is not bedtime-driven.

1. Summary Table — Night of 2026-08-18

Metric	Value
Bedtime (session start)	23:09 on 08-18 (3rd consecutive on-schedule)
Time in bed	6.17 h (370.3 min)
Total sleep duration	5.95 h (356.8 min)
Sleep efficiency (TST/TIB)	96.34% — rank 18/31; +1.2 pts vs. 7-day (95.18%), +0.9 pts vs. 30-day (95.42%)
Awake time	13.55 min — rank 14/31; −31% vs. 7-day (19.5), −28% vs. 30-day (18.8)
Core sleep	4.31 h (258.4 min) — 72.43% (elevated; typical band 55–65%)
Deep sleep	0.53 h (31.6 min) — 8.86% — rank 4/31 lowest; −5.35 pts vs. 08-17 (14.21%); −5.70 pts vs. 30-day (14.56%)
REM sleep	1.11 h (66.7 min) — 18.71% — rank 3/31 lowest; −6.64 pts vs. 08-17 (25.35%); −5.38 pts vs. 30-day (24.09%)
CPAP usage	6.12 h — −8.5% vs. 7-day (6.69), −8.8% vs. 30-day (6.71)
AHI	1.63 — rank 20/31; −5% vs. 7-day (1.71), essentially at 30-day (1.64); rebounded +47% vs. 08-17 (1.11)
RDI	1.80 (RERA = 1)
OA / H / CA / UA / RERA	7 / 2 / 1 / 0 / 1 — CA and RERA both return after 08-17's clean profile
OA per hour	1.14 /h — matches 30-day OA-rate (~1.14/h); +43% vs. 08-17 (0.80/h)
Pressure avg	7.55 cmH₂O — essentially flat vs. 08-17 (7.57); still below 7-day (7.72) and above 30-day (7.44)
Pressure 90th %ile	11.11 cmH₂O — +1.57 vs. 08-17 (9.54); rank 29/31 (near-top); +0.93 vs. 7-day (10.18), +1.11 vs. 30-day (10.00)
Leak avg	1.19 L/min — 4th consecutive uptick; 2nd consecutive night ≥1.0; +138% vs. 7-day (0.50)
Leak 90th %ile	3.6 L/min — +2.4 vs. 08-17 (1.2); tail is now expanding, not just baseline; rank 26/31
Leak max	22.8 L/min
Compliance flag	1 (compliant)

2. 10-Night Trend & Context

Date	AHI	OA	H	CA	RERA	Leak avg	Leak 90	P_avg	P_90	Hours	Deep %	REM %	Awake (min)	Eff
08-08	2.34	6	10	0	0	0.87	3.6	7.63	9.90	6.83	16.17	26.60	15.6	96.12%
08-09	1.34	6	3	0	2	3.00	10.8	7.79	10.94	6.72	10.97	23.65	15.1	96.19%
08-10	0.58	3	0	0	0	0.17	0.0	6.89	9.46	5.20	15.30	20.31	20.1	94.73%
08-11	2.57	12	4	0	0	0.42	1.2	7.28	10.02	6.62	13.57	25.24	21.1	94.61%
08-12	2.68	13	3	2	1	0.28	1.2	7.99	10.00	6.72	20.78	19.11	28.1	92.80%
08-13	1.29	8	0	2	0	0.20	1.2	7.01	9.54	7.75	10.93	26.70	20.1	95.60%
08-14	1.63	11	1	0	1	0.22	0.0	7.94	10.86	7.38	9.73	29.20	33.6	92.19%
08-16	1.13	3	3	1	2	0.91	4.8	7.92	10.94	6.22	14.73	19.78	6.0	98.34%
08-17	1.11	5	2	0	0	1.14	1.2	7.57	9.54	6.28	14.21	25.35	7.5	97.95%
08-18	1.63	7	2	1	1	1.19	3.6	7.55	11.11	6.12	8.86	18.71	13.6	96.34%

(08-15 not in table: no matched Watch sleep-stage session — CPAP-only that night was AHI 1.53, OA 8, Leak 0.31, P_avg 8.35, Hours 5.88.)

Three-night composite (08-16 + 08-17 + 08-18): AHI mean 1.29 • OA total 15 across 18.62 h use = 0.81 OA/h • Awake total 27.1 min across 18.30 h TIB = 2.47% awake fraction • Deep mean 12.60% • REM mean 21.28% • Efficiency mean 97.54%. Compared to the tight 2-night composite through 08-17 (Deep 14.47%, AHI 1.12), tonight's addition drags Deep down ~2 pts and AHI up ~0.17. Awake time and efficiency stay strong; the architecture channel weakened more than the respiratory channel.

Seven-night 08-12 → 08-18 obstructive composite: OA 59, H 12, CA 6, RERA 5 across 46.35 h = 1.27 OA/h. Compare to the 6-night 08-12 → 08-17 composite (48 OA in 40.23 h = 1.19 OA/h). One night's regression added 7 OA and pushed the trailing rate up 7%. The cluster is not yet closed.

Leak vs. sleep architecture — 31 matched nights (through 08-18)

Buckets on nightly leak average:

Low-leak (≤ 0.30 L/min): n = 7 — Mean deep: 13.27% | Mean awake: 23.3 min
Higher-leak (> 0.30 L/min): n = 24 — Mean deep: 14.69% | Mean awake: 17.3 min
Very-high-leak (≥ 1.0 L/min): n = 9 — Mean deep: 13.44% | Mean awake: 13.4 min

the user r (leak avg vs. deep %): −0.06 (n=31). the user r (leak 90 vs. deep %): −0.10. the user r (leak avg vs. awake min): −0.30 (unchanged directional signal — leak-noisy nights coincide with slightly more consolidated sleep). the user r (AHI vs. deep %): +0.04. the user r (AHI vs. awake min): −0.09. the user r (OA vs. awake min): +0.12. the user r (leak avg vs. REM %): +0.21. the user r (pressure avg vs. deep %): −0.02. the user r (pressure 90th vs. deep %): +0.05.

Interpretation: In the now-31-night matched sample, none of the CPAP-side metrics have a meaningful directional relationship with next-morning deep-sleep %. Tonight's low deep does not fit a CPAP-driven explanation — leak avg 1.19 (very-high bucket) has a mean deep of 13.44% historically, so tonight's 8.86% is a large downward outlier within its own bucket. Only three of the six lowest-deep nights (08-01 5.82%, 07-20 8.31%, 08-04 8.54%, 08-18 8.86%, 08-14 9.73%, 07-25 9.91%) have leak avg > 1.0. The signal that architecture compression is driven by non-CPAP factors is now strengthened, not weakened, by adding tonight.

3. Comparison to Trailing Averages

Metric	Tonight (08-18)	7-day avg (matched, through 08-17)	30-day avg (matched, through 08-17)
AHI	1.63	1.71	1.64
OA count	7	8.57	7.47
OA per hour	1.14 /h	~1.28 /h	~1.14 /h
Leak avg (L/min)	1.19	0.50	0.80
Leak 90th (L/min)	3.6	1.54	2.44
Pressure avg (cmH₂O)	7.55	7.72	7.44
Pressure 90th (cmH₂O)	11.11	10.18	10.00
Usage hours	6.12	6.69	6.71
Deep sleep %	8.86%	14.18%	14.56%
REM sleep %	18.71%	23.67%	24.09%
Awake time (min)	13.6	19.5	18.8
Efficiency	96.34%	95.18%	95.42%
TIB (h)	6.17	6.62	6.63

Verdict: mixed night with the split flipped from the 08-16/17 pattern. Sleep consolidation (awake, efficiency) is still better than baseline, but architecture is decisively worse — deep is −5.7 pts and REM is −5.4 pts vs. 30-day, both bottom-4 of matched history. On the CPAP side: leak avg and leak 90 both blew past their trailing averages, and P_90 climbed to a top-3 (worst) reading; AHI and OA/h are on par with 30-day but rebounded meaningfully from 08-17's low. This is the mirror image of 08-16/17, where every channel except leak avg outperformed baseline.

Weekly AHI/Leak/Pressure/OA trend (CPAP-only, full week nights)

Week (Wed–Tue)	AHI avg	Leak avg	P_avg	P_90 avg	OA avg	Hours	n
07-16 to 07-21	1.62	0.46	8.18	10.78	6.7	6.38	6
07-22 to 07-28	1.68	0.87	7.41	10.22	8.6	6.76	7
07-29 to 08-04	1.74	0.70	7.03	9.27	7.3	7.12	7
08-05 to 08-11	1.83	1.16	7.51	10.13	7.6	6.37	7
08-12 to 08-18 (final, n=7)	1.57	0.61	7.76	10.34	7.9	6.62	7

Full-week close: (1) weekly AHI 1.57 — the lowest of the last 5 weeks (prior lowest: 07-16 to 07-21 at 1.62). Even with tonight's rebound, the week ends favorably vs. the 4-week rising trend of 1.62 → 1.68 → 1.74 → 1.83. (2) Weekly OA 7.9 — off its 08-05 → 08-11 peak (7.6) actually rose here, driven by the 25 OA count across 08-11/12. (3) Weekly leak avg 0.61 — down from the prior week's 1.16, but tonight's 1.19 lifted it from the 6-night partial of 0.51. (4) Weekly P_avg 7.76 / P_90 10.34 — pressure demand up modestly on the week, with tonight's P_90 = 11.11 being the top contributor.

The 5-week AHI trend (1.62 → 1.68 → 1.74 → 1.83 → 1.57) is a meaningful bend down after four weeks of climbing. The 2-night improvement of 08-16/17 carried the week; tonight's regression to 1.63 was insufficient to spoil the weekly picture. This is the first confirmed weekly-level AHI improvement in over a month.

4. Outlook

The obstructive cluster is neither closed nor reopened — call it "resolving unevenly." AHI/OA rebounded tonight (1.11 → 1.63; 5 → 7) but stayed under the reopen thresholds (AHI 2.5, OA 10). The two-night clean streak became a 2-of-3 pattern, and CA + RERA both reappeared. The auto-titrator's P_90 jump (+1.57) is the machine acknowledging the load returned. Interpret this as: the 08-11 → 08-15 elevated period has not fully cleared, and the recovery pattern is intermittent rather than monotonic. Watch 08-19: another sub-1.5 AHI reopens the "cluster closing" thesis; a bounce ≥ 2.5 puts it back on the table.

Leak channel — decision point on 08-19. Four nights of rising leak avg (0.31 → 0.91 → 1.14 → 1.19), two consecutive at ≥1.0, and tonight the tail (Leak_90) expanded from 1.2 to 3.6. The "persistent small-magnitude drift" interpretation from 08-17 no longer fits cleanly — tonight looks like baseline elevation plus return of spikes. Per the 08-17 report's threshold, one more night at ≥1.0 crosses the 3-consecutive mark that warrants a physical mask/strap/cushion check. Do the check pre-emptively tonight if convenient — it costs a minute and would head off the trend regardless of 08-19's data.

Sleep architecture — this is the most concerning channel tonight. Deep 8.86% and REM 18.71% are both bottom-4 of the entire matched history, on a night with on-schedule bedtime and no obvious CPAP-side driver. Core-sleep dominance (72.4%) means the Watch is scoring a lot of the night as light sleep — this can happen if slow-wave and REM cycles are shortened or disrupted below the classifier's confidence threshold. Possible non-CPAP contributors to consider (not diagnosable from this data alone): alcohol proximate to bedtime, late caffeine, evening exercise, illness onset, elevated body temperature, or medication changes. Two-night composite deep (11.54%) now materially below the 30-day baseline (14.56%). If 08-19 also comes in with Deep < 12%, treat as a 3-night architecture regime worth investigating for a behavioral or health explanation. If 08-19 recovers to 14%+, tonight was a one-night noise event on top of the 08-13/14 dip.

What to watch on 08-19. (1) Deep % — recovery to 12%+ vs. another sub-12% night is the biggest single signal in this dataset right now. (2) Leak avg — third consecutive ≥1.0 triggers the mask-check action item; a drop to <0.5 closes it. (3) AHI/OA direction — sub-1.5 = cluster resolution back on; ≥2.5 = reopen. (4) P_90 — sustained ≥11 for a second night would be the auto-titrator saying load is not receding. (5) REM stability — tonight is 3rd-lowest REM of matched history; a second sub-20% REM night would push toward a genuine REM-suppression pattern.

5. Caveats

CPAP/Watch alignment tonight is very tight (6.12 h vs 6.17 h, Δ 3.1 min) — the closest alignment of the trailing three nights. Per-hour rates and architecture comparisons are directly interpretable.
The low-deep / low-REM combination is what makes tonight distinct. Individually, either could be Watch classifier noise on any given night. Together they suggest genuine architecture compression rather than a mis-classification.
Core-sleep percentage of 72.43% is well outside the typical 55–65% band and is what "picks up the slack" when both deep and REM under-score. Watch's Core category functionally aggregates N1+N2 and any lower-confidence stage — a high Core % often flags shallower / more fragmented cycling rather than more sleep.
Leak trend interpretation shifted from 08-17. Yesterday: "baseline-dominated, low tail — probably small-magnitude drift." Tonight: "baseline sustained AND tail expanded." Two different signatures. The 08-19 data point is disproportionately important here — a third consecutive ≥1.0 night with an expanding tail is a strong mask-fit signal; a drop under 0.5 makes the 4-night rising streak look like a limited episode.
AHI 1.63 rebound is not statistically significant on its own — it's inside the trailing-30-day mean (1.64). The narrative weight comes from the comparison to the immediately-preceding 2 nights (1.13, 1.11), not to the broader baseline.
P_90 = 11.11 is the notable pressure signal, not P_avg (which barely moved). This is the auto-titrator responding to specific events in the night rather than raising its overall demand — indicating event clusters within the session rather than sustained high load.
No SpO2 data in tonight's CPAP CSV export (columns present but blank) — consistent with every export in the current format.
CPAP-nights CSV format. Newer files include 90th-percentile leak; older files use 95th-percentile. Both treated as the same "tail leak" metric for trend purposes.
Matched-night count on this run: n=31. CPAP-only nights (no Watch match) in the last 6 weeks remain: 07-17, 08-07, 08-15. Weekly CPAP-only rollups use full n=6–7; correlation/matched stats use n=31.
Correlation coefficients (n=31) essentially unchanged from prior report's n=30 — near-zero across all CPAP-vs-architecture pairings. Tonight fits the pattern that architecture is driven by non-CPAP factors.
Session-start convention. The date field of the sleep session (2026-08-18T23:09:57) marks the start calendar date, which lines up with the OSCAR "Period" for 2026-08-18. This convention has held consistently through the matched-history window.

See all past reports →