Sleep & CPAP Report — Night of 2026-07-09
Sources: sleep_export_complete.json (Apple Watch sleep stages) + oscar_data.csv (ResMed AirSense/AirFlow 11 via OSCAR). Matched by calendar date of sleep-session start (22:37:26 on 07-09).
Flags
- Leak spike, isolated but sharp. Leak_Avg jumped to 1.57 L/min (95th %ile: 6.0 L/min) last night — 4.4x the trailing-7-day average of 0.36 L/min, and the highest leak reading since 2026-07-01 (6.22 L/min). It follows 6 straight nights (07-02 to 07-08) with Leak_Avg under 0.3.
- Deep sleep collapsed the same night. Deep sleep was 10.64% of total sleep — the lowest of the trailing 7 nights (avg 21.21%) and below the trailing 30-day average (17.92%). This is consistent with a broader pattern in the data: across all matched nights, high-leak nights (Leak_Avg > 1.5, n=6) average 11.65% deep sleep vs 20.25% on low-leak nights (n=13) — a 43% relative reduction.
- Standing pattern, not new: AHI roughly tripled vs. three weeks ago. Week-over-week AHI: 06-15–06-22 avg 0.87 → 06-23–06-29 avg 3.41 → 06-30–07-06 avg 2.80 → 07-07–07-09 avg 2.75. AHI has stayed 3-4x baseline since 06-23 and has not returned to early-window levels. Worth a look if this continues.
- Prior 4-night leak streak (06-24 to 06-27): Leak_Avg exceeded 2.0 L/min every night for 4 consecutive nights (2.85, 3.26, 2.52, 2.92), the only such streak in the dataset. That block also had the lowest deep-sleep readings in the record (9.7%, 12.5%, 8.7%, 13.1%) — worth remembering as context even though it predates last night.
- Last night's CPAP pressure averaged 7.18 cmH₂O, the lowest of the entire 25-night record (30-day avg 5.97 cmH₂O is skewed by early zero-pressure readings — see caveats). Within the matched 7-day window, pressure had been running 8.0-9.3 cmH₂O, so the leak spike coincided with the mask running at lower, not higher, pressure — points toward a mask-seal/positioning issue rather than a pressure-driven leak.
1. Summary table — night of 2026-07-09
| Metric |
Value |
| Bedtime (session start) |
22:37 |
| Time in bed |
6.32 h (6h 19m) |
| Total sleep duration |
5.90 h (5h 54m) |
| Sleep efficiency (TST/TIB) |
93.3% |
| Awake time |
25.6 min |
| Core sleep |
3.75 h — 63.6% |
| Deep sleep |
37.6 min — 10.6% |
| REM sleep |
91.3 min — 25.8% |
| CPAP usage hours |
6.62 h |
| AHI |
2.12 events/hr |
| RDI |
2.12 events/hr |
| Obstructive apneas (OA) |
13 |
| Hypopneas (H) |
0 |
| Central apneas (CA) |
1 |
| Unclassified/other (UA) |
0 |
| RERA |
0 |
| Pressure avg / 95th |
7.18 / 9.52 cmH₂O |
| Leak avg / 95th |
1.57 / 6.00 L/min |
2. CPAP–sleep-stage correlations
Across the 19 nights with both a matched CPAP record and a full sleep-stage breakdown (i.e., excluding short/nap sessions and 5 CPAP nights the Watch didn't capture a full sleep session for — see caveats):
- AHI vs. Deep-sleep %: r ≈ −0.08 — essentially no linear relationship in this dataset. Obstructive event rate alone doesn't predict how much deep sleep a given night has.
- AHI vs. awake minutes: r ≈ +0.41 — a moderate positive relationship. Nights with more apnea/hypopnea events tend to have more time coded awake, as expected (events often trigger brief arousals).
- Leak vs. awake minutes: r ≈ +0.19 — weak positive relationship; leak alone is a noisier predictor of arousal than AHI is.
- Leak vs. deep sleep (grouped): the clearest signal in the data. High-leak nights (Leak_Avg > 1.5 L/min) average 11.65% deep sleep vs. 20.25% on low-leak nights — the largest effect size of anything tested here, larger than AHI's effect. This lines up with both last night and the 06-24–06-27 streak.
Read these as descriptive, not causal — n is small (19 matched nights) and leak/AHI/pressure move together on several nights, so it's hard to isolate one driver.
3. Trend comparison
Last night vs. trailing 7-day average (07-02 to 07-08):
| Metric |
Last night (07-09) |
7-day avg |
Δ |
| AHI |
2.12 |
2.31 |
−0.19 (slightly better) |
| Leak avg |
1.57 |
0.36 |
+1.21 (4.4x) |
| Pressure avg |
7.18 |
8.49 |
−1.31 |
| Usage hours |
6.62 |
6.43 |
+0.19 |
| Deep sleep % |
10.64% |
21.21% |
−10.57 pts |
| REM sleep % |
25.82% |
24.00% |
+1.82 pts |
| Awake min |
25.6 |
13.05 |
+12.55 (~2x) |
Last night vs. trailing 30-day average (06-15 to 07-08, 23 CPAP nights / ~16 matched sleep nights):
| Metric |
Last night |
30-day avg |
Δ |
| AHI |
2.12 |
2.42 |
−0.30 |
| Leak avg |
1.57 |
1.05 |
+0.52 |
| Usage hours |
6.62 |
6.27 |
+0.35 |
| Deep sleep % |
10.64% |
17.92% |
−7.28 pts |
| Awake min |
25.6 |
15.89 |
+9.71 |
Net read: last night was CPAP-effective by AHI standards (in line with or better than recent norms, and 0 hypopneas), but the leak spike and near-doubled awake time both point to a disrupted, lighter-quality night — mainly evident in the collapsed deep-sleep share rather than in the apnea count itself.
Caveats / data notes
- Sleep-stage data has no per-epoch timestamps in this export — only aggregate durations/percentages per session plus total awake time. "Transitions" are inferred from aggregate awake time and stage share, not from a stage-by-stage timeline.
- 5 CPAP nights (06-15, 06-16, 06-18, 06-20, 07-03) have no matching Watch sleep session with stage data (either a nap-length record or none at all), so they're excluded from the correlation and 7-day-sleep-stage figures.
- 06-21 has no CPAP record at all (mask likely not used or file gap) — excluded from all CPAP trend figures.
- Matching CPAP "Period" date to Watch session start date was validated by minimizing the gap between CPAP usage hours and Watch time-in-bed across all 22 matched pairs (median gap ~0.17h); this is a best-fit heuristic, not a guaranteed 1:1 match from raw timestamps.
- 30-day Pressure_Avg (5.97) is pulled down by 6 early nights (06-15–06-22) recorded as 0.00 cmH₂O, which looks like a logging gap rather than a true 0; treat the 30-day pressure comparison as unreliable and lean on the 7-day figure instead.