Sources: health-report-2026-07-26-to-2026-07-27.json (Apple Watch sleep stages, exported 05:48 on 07-27) + OSCAR_Greg_by_Day_2026-07-26.csv (ResMed AirSense/AirFlow 11 via OSCAR). Matched by calendar date of sleep-session start (22:57 on 07-26).
The leak spike resolved and AHI/OA fully reverted to baseline as predicted — but awake time and sleep efficiency did not follow, breaking the simple leak-driven story. Leak avg fell 77% (1.69 → 0.39 L/min, leak 90th 7.2 → 0.0) and AHI/OA landed back inside the pre-spike range (AHI 1.24, OA 5 — both within the 07-20–23 baseline of 0.88–1.59 / 2–7). Yet awake time only eased from 30.6 to 26.6 min and efficiency was flat (92.5% → 92.9%). 07-25 and 07-26 are now the 3rd- and 5th-highest awake-time nights, and the 3rd- and 4th-lowest efficiency nights, of all 35 tracked nights — a two-night streak in the dataset's worst decile on both metrics, even though the CPAP-side metrics (leak, AHI) that supposedly drive this normalized on night two.
This divergence pulled the leak/AHI-vs-awake correlations down, not up. Adding tonight (low leak, low AHI, high awake) as a counterexample: r(leak avg, awake min) fell from +0.12 to +0.096 and r(AHI, awake min) fell from +0.18 to +0.138 (n=35). The leak-vs-deep-sleep relationship held essentially flat (−0.30 → −0.298). Net effect: tonight weakens confidence that leak/AHI alone explain the awake-time pattern — something else (possibly a one-night lag effect from 07-25's disruption, or an unmeasured factor) may be contributing.
Deep sleep partially recovered (9.9% → 14.3%) but remains below both trailing averages (7-day 16.4%, 30-day 16.6%) — the third night in the last four below its own recent baseline (07-23: 25.5%, 07-24: 16.7%, 07-25: 9.9%, 07-26: 14.3%).
| Metric | Value |
|---|---|
| Bedtime (session start) | 22:57 |
| Time in bed | 6.28 h (376.8 min) |
| Total sleep duration | 5.84 h (350.2 min) |
| Sleep efficiency (TST/TIB) | 92.9% (flat vs. 07-25, still below 7-day avg) |
| Awake time | 26.6 min ↓ from 30.6 but 5th-highest in dataset |
| Core sleep | 3.54 h (212.3 min) — 60.6% |
| Deep sleep | 0.84 h (50.2 min) — 14.3% ↑ from 9.9% but still below trailing avg |
| REM sleep | 1.46 h (87.8 min) — 25.1% |
| CPAP usage | 6.47 h |
| AHI | 1.24 ↓ (back within 07-20–23 baseline) |
| RDI | 1.24 |
| OA / H / CA / UA / RERA | 5 / 3 / 0 / 0 / 0 |
| Pressure avg | 7.59 cmH₂O |
| Pressure 90th %ile | 10.02 cmH₂O |
| Leak avg | 0.39 L/min ↓↓ (resolved, down 77% from 1.69) |
| Leak max | 14.4 L/min |
| Leak 90th %ile | 0.0 L/min ↓ (down from 7.2) |
| Date | AHI | OA | RERA | Leak avg | Leak 90th | Deep % | Awake (min) | Efficiency |
|---|---|---|---|---|---|---|---|---|
| 07-19 | 1.29 | 6 | 1 | 0.92 | 0.0 | 10.1% | 14.0 | 96.2% |
| 07-20 | 0.94 | 2 | 0 | 0.03 | 0.0 | 8.3% | 21.1 | 94.4% |
| 07-21 | 0.88 | 4 | 1 | 0.85 | 1.2 | 28.4% | 27.1 | 93.3% |
| 07-22 | 1.59 | 7 | 1 | 1.08 | 2.4 | 12.0% | 10.0 | 97.5% |
| 07-23 | 1.29 | 7 | 0 | 0.48 | 2.4 | 25.5% | 13.0 | 96.8% |
| 07-24 | 2.34 | 15 | 0 | 0.48 | 0.0 | 16.7% | 24.6 | 93.9% |
| 07-25 | 1.72 | 10 | 0 | 1.69 | 7.2 | 9.9% | 30.6 | 92.5% |
| 07-26 | 1.24 | 5 | 0 | 0.39 | 0.0 | 14.3% | 26.6 | 92.9% |
Split at leak ≤ 0.30 vs. > 0.30 L/min (07-26's leak of 0.39 falls just above the threshold, into the higher-leak bucket, despite being a comparatively mild reading):
the user r (leak avg vs. deep %): −0.298 (n=35, was −0.30). the user r (leak avg vs. awake min): +0.096 (was +0.12). the user r (AHI vs. awake min): +0.138 (was +0.18). the user r (AHI vs. deep %): +0.056 (was +0.05).
Interpretation: The leak→deep-sleep relationship held steady. But both awake-time correlations weakened with tonight's data point added, because 07-26 combines low leak (0.39) and low AHI (1.24) with high awake time (26.6 min, 5th-highest on record) — a genuine counterexample to the "leak/AHI drives awake time" story that the prior two reports had been building. All |r| values remain in the weak 0.05–0.30 range; treat directional tendencies, not tonight's outcome, as the reliable signal.
| Metric | Tonight (07-26) | 7-day avg (07-20–07-26) | Prior 7-day avg (07-13–07-19) | 30-day avg (06-27–07-26) |
|---|---|---|---|---|
| AHI | 1.24 | 1.43 | 2.22 | 2.45 |
| Leak avg (L/min) | 0.39 | 0.71 | 0.42 | 0.82 |
| Pressure avg (cmH₂O) | 7.59 | 7.46 | 8.77 | 8.41 |
| Usage hours | 6.47 | 6.77 | 6.32 | 6.58 |
| Deep sleep % | 14.3% | 16.4% | 14.7% | 16.6% |
| REM sleep % | 25.1% | 26.4% | 21.5% | 24.2% |
| Awake time (min) | 26.6 | 21.9 | 15.8 | 16.9 |
AHI, leak, and pressure are all now at or below their trailing 7-day and 30-day averages — the CPAP-side picture looks clean. Deep sleep (14.3%) sits about 2 points below both trailing windows, and awake time (26.6 min) is roughly 1.2x the 7-day average and 1.6x the 30-day average — the sleep-architecture side is the lagging piece tonight, not the therapy metrics.
| Week | AHI avg | Leak avg | n |
|---|---|---|---|
| 06-22 to 06-28 | 2.84 | 1.98 | 7 |
| 06-29 to 07-05 | 3.24 | 1.41 | 7 |
| 07-06 to 07-12 | 3.00 | 0.43 | 7 |
| 07-13 to 07-19 | 2.22 | 0.42 | 7 |
| 07-19 to 07-25 (prior trailing 7d) | 1.44 | 0.79 | 7 |
| 07-20 to 07-26 (current trailing 7d) | 1.43 | 0.71 | 7 |
The multi-week AHI downtrend holds essentially flat at its recent low (1.44 → 1.43). Trailing leak eased back down (0.79 → 0.71) as the 07-25 spike aged out of most of the window's influence — it's still the single highest value inside the current 7-day window, but the six other nights (0.03–1.24) keep the average moderate.
CPAP-side metrics (AHI, OA, leak, pressure): all clear. No action items — this looks like a fully resolved single-night event on 07-25, exactly as anticipated in the prior report.
Primary open question: why didn't awake time and efficiency follow leak/AHI back down? Two consecutive nights (07-25, 07-26) now rank in the dataset's worst decile for both awake time and efficiency, despite CPAP metrics being unremarkable on the second of those nights. Possible explanations worth considering over the next few nights: a one-night lag/carryover effect from a disrupted prior night, an unmeasured factor (caffeine, alcohol, stress, room temperature, etc.) coinciding with both nights, or simply two independent rough nights. If 07-27 also shows elevated awake time or low efficiency with otherwise-normal CPAP metrics, that would point away from CPAP therapy as the driver and toward something else in the routine around 07-25–07-26.
Deep sleep: Trending back toward baseline (9.9% → 14.3%) but not there yet; watch for continued recovery toward the ~16–17% recent norm.
oscar_data.csv) and are excluded from pressure-based averages/trends for that period.