Sources: health-report-2026-08-11-to-2026-08-12.json (Apple Watch sleep stages, exported 07:48 on 08-12) + OSCAR_Greg_by_Day_2026-08-11.csv (ResMed AirSense/AirFlow 11 via OSCAR). Session start (23:36 on 08-11) is on the same calendar date as the CPAP Period, so the match is direct. CPAP usage (6.62 h) is close to Apple Watch time in bed (6.51 h) — well-aligned tonight, no meaningful gap.
| Metric | Value |
|---|---|
| Bedtime (session start) | 23:36 on 08-11 |
| Time in bed | 6.51 h (390.9 min) |
| Total sleep duration | 6.16 h (369.8 min) |
| Sleep efficiency (TST/TIB) | 94.6% — rank 17/55 from low (below median 96.1%; awake-time drag) |
| Awake time | 21.1 min — above 7-day (16.1) and 30-day (17.9) averages |
| Core sleep | 3.77 h (226.3 min) — 61.19% |
| Deep sleep | 0.84 h (50.2 min) — 13.57% — rank 20/55, below 30-day avg (14.91%) |
| REM sleep | 1.56 h (93.3 min) — 25.24% — above 7-day (23.7%) and 30-day (24.2%) |
| CPAP usage | 6.62 h — matched to dataset median; rebounded from 5.20 last night |
| AHI | 2.57 — above 7-day (1.91) and 30-day (1.81); 4.4x last night's 0.58 |
| RDI | 2.57 |
| OA / H / CA / UA / RERA | 12 / 4 / 0 / 1 / 0 — OA count is 3rd-highest of trailing 8 nights |
| Pressure avg | 7.28 cmH₂O — near 7-day (7.52) and 30-day (7.69) |
| Pressure 90th %ile | 10.02 cmH₂O |
| Leak avg | 0.42 L/min — 2nd clean night in a row; below 30-day baseline (0.77) |
| Leak 90th %ile | 1.2 L/min — normal band |
| Compliance flag | 1 (compliant) |
| Date | AHI | OA | H | CA | RERA | Leak avg | Deep % | Awake (min) | Efficiency |
|---|---|---|---|---|---|---|---|---|---|
| 08-04 | 3.07 | 14 | 5 | 2 | 1 | 0.43 | 8.5% | 35.6 | 91.6% |
| 08-05 | 1.49 | 7 | 2 | 1 | 0 | 1.90 | 11.3% | 10.5 | 97.3% |
| 08-06 | 2.83 | 11 | 4 | 2 | 1 | 0.56 | 15.9% | 9.5 | 97.3% |
| 08-07 | 1.69 | 8 | 2 | 1 | 1 | 1.20 | 20.5% | 6.5 | 98.3% |
| 08-08 | 2.34 | 6 | 10 | 0 | 0 | 0.87 | 16.2% | 15.6 | 96.1% |
| 08-09 | 1.34 | 6 | 3 | 0 | 2 | 3.00 | 11.0% | 15.1 | 96.2% |
| 08-10 | 0.58 | 3 | 0 | 0 | 0 | 0.17 | 15.3% | 20.1 | 94.7% |
| 08-11 | 2.57 | 12 | 4 | 0 | 1 UA | 0.42 | 13.6% | 21.1 | 94.6% |
Buckets on nightly leak average:
the user r (leak avg vs. deep %): −0.26 (n=55). the user r (leak 90th vs. deep %): −0.28. the user r (leak avg vs. awake min): −0.04. the user r (AHI vs. deep %): +0.01. the user r (AHI vs. awake min): +0.04.
Interpretation: Tonight is a clean split-test night for the leak vs. AHI question. Leak sits in the higher-but-not-high band (0.42) yet AHI spiked to 2.57 with 12 obstructive events — reinforcing what the r-values already say: AHI's night-to-night swings are essentially uncorrelated with leak or with sleep-stage disruption in the historical data, while leak retains a persistent weak-negative relationship with deep-sleep %. In other words, tonight's respiratory regression looks driven by something other than the mask — positional, congestion, alcohol, sleep-onset physiology, or normal biological variance. The deep-sleep dip (13.57%) is consistent with the very-high-leak bucket's mean of 13.5%, but leak wasn't actually elevated tonight, so the driver of the deep dip is more likely the elevated awake time (21.1 min) than any respiratory event pathway.
| Metric | Tonight (08-11) | 7-day avg (08-04–08-10) | 30-day avg (through 08-10) |
|---|---|---|---|
| AHI | 2.57 | 1.91 | 1.81 |
| Leak avg (L/min) | 0.42 | 1.16 | 0.77 |
| Leak 90th (L/min) | 1.2 | 3.94 | 2.28 |
| Pressure avg (cmH₂O) | 7.28 | 7.52 | 7.69 |
| Usage hours | 6.62 | 6.40 | 6.66 |
| Deep sleep % | 13.57% | 14.11% | 14.91% |
| REM sleep % | 25.24% | 23.71% | 24.18% |
| Awake time (min) | 21.1 | 16.1 | 17.9 |
Mixed night: leak and pressure metrics stay well below trailing norms (leak avg 63% below the 7-day, leak 90th 70% below), usage is back at baseline, REM beats trend — but AHI, awake time, and deep % all miss trend on the wrong side. The AHI miss is the sharpest (+35% vs. 7-day, +42% vs. 30-day) and is entirely accounted for by the OA jump; hypopneas (4) and centrals (0) are in-line.
| Week | AHI avg | Leak avg | n |
|---|---|---|---|
| 06-17 to 06-23 | 1.24 | 0.24 | 6 |
| 06-24 to 06-30 | 3.74 | 1.99 | 7 |
| 07-01 to 07-07 | 2.66 | 1.22 | 7 |
| 07-08 to 07-14 | 2.78 | 0.52 | 7 |
| 07-15 to 07-21 | 1.82 | 0.43 | 7 |
| 07-22 to 07-28 | 1.68 | 0.87 | 7 |
| 07-29 to 08-04 | 1.74 | 0.70 | 7 |
| 08-05 to 08-11 (complete) | 1.83 | 1.16 | 7 |
The current week closes at AHI 1.83, Leak 1.16 — essentially in-line with the last four weeks on AHI (all in the 1.68–1.83 band) but the highest weekly leak average in six weeks, driven almost entirely by the 08-05 (1.90), 08-07 (1.20), and 08-09 (3.00) spikes. The two most recent nights (08-10, 08-11) both landed in the low-leak band, so if the next week continues at 0.4–0.5 the running weekly leak average will drop back toward the 0.4–0.9 range seen through most of July and early August.
Regression to the mean, not a new problem. Last night's dataset-best AHI (0.58) was a beautiful outlier; tonight's 2.57 is more representative of the week and the last four weeks generally. The specific texture of tonight's respiratory events is worth flagging — 12 OAs is on the high end of what's been logged recently, and the fact that it happened with leak fully resolved argues against a mask-fit driver. Watch for whether the next 2–3 nights cluster at OA ≥ 10; a run of that would suggest either a positional shift (more supine sleep), congestion/allergy, or that the current pressure profile (avg 7.28, 90th 10.02) is at times undertreating obstructive events. If the OA count normalizes to the 4–8 range over the coming nights, tonight is just noise. The leak resolution is real and holding — 2 nights running in the low-leak band with no meaningful excursions — so the previous week's leak issue can be considered resolved unless a new spike appears. Deep-sleep architecture remains the softer story: 4 of the last 8 nights below 14%, dragging the 7-day average to 14.11% (below the 30-day 14.91%). Not urgent, but if the trend continues another week it moves from "watch" to "flag."
oscar_data.csv / newer per-day CSVs both feeding the merge — not data errors.