🔴 Leak_Max 40.80 L/min — new 26-night window RECORD (z = +2.02 vs 30d). Previous window high was 2026-09-22's 36.0; tonight exceeds it by +13.3%. The 30d mean Leak_Max is 23.12 (sd 8.76), so 40.80 is a +2-sigma outlier. In the entire 103-night confident set (post-2026-06-23), this ranks #4 all-time behind only 06-30 (120.0), 07-01 (116.4), and 06-26 (42.0) — all three of those are early-period readings whose channels are flagged as low-confidence in the caveats convention. Among post-low-confidence nights tonight is the single highest Leak_Max on record. Yesterday's "streak partially broken" verdict is now fully reversed: not only did Leak_Max rebound above the 22.8 cluster-threshold (as noted yesterday), it blew through the window top. Combined with Leak_Avg 2.18 (rank 4/26, z = +1.29, +76.8% vs 30d — 10th-highest Leak_Avg in the 103-night confident set) and Leak_90 8.40 (rank 6/26, z = +1.22, +98.1% vs 30d), the leak signal is intensifying across every leak dimension except brief-peak-only.
🔴 FL_P95 0.14 — TIES 10-10's 26-night window record (z = +2.10 vs 30d). The 10-11 report concluded yesterday's reversion to FL_P95 0.09 meant the 10-10 spike was "a single-night event, not the start of a sustained trend." Tonight contradicts that verdict: FL_P95 is back at 0.14, tied with 10-10 for the window top. In the full 26-night window only 2 nights (10-10, 10-12) have reached 0.14; prior to 10-10 the window high was 0.13 (10-04). The 4-night sustained-FL pattern is now spike / reversion / spike (10-10: 0.14 → 10-11: 0.09 → 10-12: 0.14), not a single-night event. MinAbv010 26.7 min (rank 3/26, z = +1.71, +49.5% vs 30d) corroborates: only 10-10 (32.0) and 09-20 (28.6) are higher in the window. FL_Avg 0.021 (z = +1.42, +28.8% vs 30d) is in the top-5 of the window. MinAbv020 5.9 min (z = +0.73, +34.2% vs 30d) is also elevated. The sustained-FL distribution has returned to near-record territory.
🔴 Pressure_Avg 8.20 — rank 3/26 (z = +1.42, +7.6% vs 30d); Pressure_90 10.74 — rank 3/26 (z = +0.81). 2nd consecutive night of elevated pressure response, but with a completely different mechanism from yesterday. Last night was a flat-avg, extreme-max profile (P_Avg 7.78, P_90 10.18, P_Max 14.60 record). Tonight is the inverse — elevated plateau, modest max (P_Avg 8.20, P_90 10.74, P_Max 11.74). Only 10-02 (8.50) and 09-20 (8.21) have higher P_Avg in the 26-night window. The 5-night pressure escalation path: P_Avg 6.89 → 7.38 → 7.38 → 7.78 → 8.20 — a monotonic climb of +1.31 cmH2O (+19%) over the week. The auto-titrator is now sitting at a durably higher operating pressure, not just reaching for brief peaks. Yesterday's prediction that P_Max 14.60 would consolidate (stay ≥ 13.5) or revert (≤ 12.5) resolves as REVERTED (11.74 is well below 12.5) — but the pressure response has shifted from peak-based to plateau-based, which is a different outcome than "transient."
🟡 Short night: TIB 335.68 min (5h 35.7m) — rank 26/26, SHORTEST sleep window in the 26-night window (z = -1.53). Usage hours 5.68 also rank 26/26 (z = -1.55); TSD 327.15 min likewise shortest. Sleep start 23:18:17 is in the normal range (within 6 minutes of yesterday's 23:24:08), so the short night came from an early wake-up at roughly 04:54 CDT, not a late sleep onset. The next-shortest TIB in the window is 09-25's 341.2 min; the mean of the 7d trailing is 387.3 min. On a per-unit-time basis this makes the FL and leak numbers more striking: the user spent 48.1 min (14.7% of sleep time) with FL ≥ 0.05 tonight versus the 30d-average 40.0 min (10.8% of sleep time) — Frac_NonZero 0.257 (within +0.7% of 30d mean 0.255) confirms this isn't a sampling-density artifact.
🟡 Deep sleep 11.04% — now a 5-night soft stretch at mean 10.66% vs 30d baseline 13.14. The pattern (10-08 → 10-12): 9.40 → 14.02 → 10.27 → 8.59 → 11.04 — every one of the last 5 nights below the 30d mean, 4 of 5 at ≤ 11%. Tonight recovered from yesterday's near-floor 8.59% (rank 2/26) and is less of an outlier (z = -0.56, rank 18/26 — only moderately below baseline), but the 5-night soft stretch is now a durable pattern, not a cluster. The 4-night slice flagged yesterday (10-08 → 10-11, mean 10.57) extends to a 5-night slice at 10.66 — nearly identical mean, now 3 straight weeks of sub-baseline deep% if the broader 7d deep% (11.95 vs 30d 13.14) continues. Deep minutes 36.13 is bolstered by the short night: on a per-minute basis the 36.13 compares to 10-11's 34.62 and 10-04's near-window-floor 32.65. The short window is not the cause of the deep deficit (deep% is time-independent); the deficit is architectural.
🟢 AHI 1.41 — 7th consecutive sub-2.0 AHI night (1.85 → 1.24 → 0.85 → 1.49 → 1.26 → 1.42 → 1.41). OA 5 (rank 23/26 — 4th-lowest in the window, z = -0.87, -42.1% vs 30d). H 3 (back to baseline, z = -0.16, rank 16/26 — tonight's event profile is NOT yesterday's hypopnea-driven one). CA 0 (tied for rank 26/26, z = -0.56), RERA 0 (tied for rank 26/26, z = -0.76). Despite all three elevated-stress signals (leak record, FL record-tie, pressure escalation), the auto-titrator held the event count down to one of the cleanest profiles in the window. The AHI-RDI gap is 0: RDI = AHI = 1.41, no RERAs tagged — the simplest, cleanest respiratory event scoring of the window. RDI 1.41 is rank 20/26 in the window (5th-lowest).
🟡 REM 21.17% — rank 21/26 (z = -0.75, -11.5% vs 30d); 10-11's 26.65% rebound was partial, not sustained. REM minutes 69.24 is a window-level sharp drop (-35.5% from yesterday's 107.38) driven partly by the short TIB. Pattern of last 5 nights' REM%: 24.30 → 20.09 → 18.96 → 26.65 → 21.17 — a volatile but net-declining profile. 7d REM% average 23.10 is now below the 30d mean 23.93 (-3.4%) for the first time in two weeks.
🟢 Sleep efficiency 97.46% — normal (z = -0.00, 30d mean 97.47). Awake time 8.53 min is at baseline (z = -0.18, 30d mean 9.62). The short night was not driven by fragmented sleep; it was driven by shorter window.
| Metric | Value |
|---|---|
| Sleep start | 2026-10-12T23:18:17-05:00 |
| Time in bed | 335.68 min (5h 35.7m) |
| Total sleep | 327.15 min (5h 27.2m) |
| Sleep efficiency | 97.46% |
| Awake time | 8.53 min |
| Core (N1+N2) sleep | 221.78 min / 67.79% |
| Deep (N3) sleep | 36.13 min / 11.04% |
| REM | 69.24 min / 21.17% |
| CPAP usage hours | 5.68 |
| Session count | 1 |
| AHI | 1.41 |
| RDI | 1.41 |
| Obstructive apneas (OA) | 5 |
| Hypopneas (H) | 3 |
| Central apneas (CA) | 0 |
| Unclassified apneas (UA) | 0 |
| RERA | 0 |
| Pressure_Avg | 8.20 |
| Pressure_Min | 4.00 |
| Pressure_Max | 11.74 |
| Pressure_90th | 10.74 |
| Leak_Avg | 2.18 |
| Leak_Max | 40.80 |
| Leak_90th | 8.40 |
| FL_Avg | 0.0210 |
| FL_P95 | 0.14 |
| FL_Max | 0.48 |
| Min_Above_005 (min) | 48.1 |
| Min_Above_010 (min) | 26.7 |
| Min_Above_020 (min) | 5.9 |
| Frac_NonZero (FL samples) | 25.70% |
| Metric | Tonight | 7d avg | 30d avg | Δ vs 7d | Δ vs 30d |
|---|---|---|---|---|---|
| AHI | 1.41 | 1.36 | 1.97 | +3.7% | -28.5% |
| RDI | 1.41 | 1.56 | 2.13 | -9.6% | -33.9% |
| OA count | 5.00 | 5.86 | 8.63 | -14.6% | -42.1% |
| H count | 3.00 | 2.86 | 3.40 | +5.0% | -11.8% |
| CA count | 0.00 | 0.14 | 0.50 | -100.0% | -100.0% |
| RERA count | 0.00 | 1.29 | 1.03 | -100.0% | -100.0% |
| Pressure_Avg | 8.20 | 7.50 | 7.62 | +9.3% | +7.6% |
| Pressure_90 | 10.74 | 10.22 | 10.20 | +5.1% | +5.3% |
| Pressure_Max | 11.74 | 12.43 | 12.10 | -5.5% | -3.0% |
| Leak_Avg | 2.18 | 1.52 | 1.23 | +43.6% | +76.8% |
| Leak_Max | 40.80 | 28.80 | 23.12 | +41.7% | +76.5% |
| Leak_90th | 8.40 | 5.14 | 4.24 | +63.3% | +98.1% |
| Usage hours | 5.68 | 6.56 | 6.42 | -13.5% | -11.5% |
| FL_P95 | 0.140 | 0.106 | 0.099 | +32.4% | +41.4% |
| FL_Avg | 0.0210 | 0.0171 | 0.0163 | +21.6% | +28.8% |
| FL_Max | 0.480 | 0.511 | 0.487 | -6.1% | -1.4% |
| Min_Above_005 | 48.1 | 42.7 | 40.0 | +12.6% | +20.2% |
| Min_Above_010 | 26.7 | 20.6 | 17.9 | +29.7% | +49.5% |
| Min_Above_020 | 5.9 | 5.66 | 4.40 | +4.3% | +34.2% |
| Frac_NonZero | 0.257 | 0.254 | 0.255 | +1.2% | +0.7% |
| Deep % | 11.04 | 11.95 | 13.14 | -7.6% | -15.9% |
| REM % | 21.17 | 23.10 | 23.93 | -8.4% | -11.5% |
| Core % | 67.79 | 64.95 | 62.94 | +4.4% | +7.7% |
| Awake min | 8.53 | 7.38 | 9.62 | +15.5% | -11.3% |
| TIB min | 335.68 | 387.30 | 380.23 | -13.3% | -11.7% |
| Total sleep min | 327.15 | 379.92 | 370.62 | -13.9% | -11.7% |
| Sleep efficiency | 97.46 | 98.07 | 97.47 | -0.6% | -0.0% |
| Metric | Recent 7d | Prior 7d | Δ | Δ % |
|---|---|---|---|---|
| AHI | 1.360 | 2.029 | -0.669 | -33.0% |
| RDI | 1.560 | 2.299 | -0.739 | -32.1% |
| OA | 5.857 | 8.857 | -3.000 | -33.9% |
| H | 2.857 | 3.571 | -0.714 | -20.0% |
| CA | 0.143 | 0.286 | -0.143 | -50.0% |
| RERA | 1.286 | 1.714 | -0.429 | -25.0% |
| Pressure_Avg | 7.501 | 7.781 | -0.280 | -3.6% |
| Pressure_90 | 10.223 | 10.414 | -0.191 | -1.8% |
| Leak_Avg | 1.519 | 1.467 | +0.051 | +3.5% |
| Leak_90 | 5.143 | 5.314 | -0.171 | -3.2% |
| Leak_Max | 28.800 | 22.629 | +6.171 | +27.3% |
| Usage hours | 6.563 | 6.414 | +0.149 | +2.3% |
| Deep % (Apple) | 11.950 | 14.093 | -2.142 | -15.2% |
| REM % (Apple) | 23.104 | 24.311 | -1.207 | -5.0% |
| Awake min (Apple) | 7.383 | 12.975 | -5.592 | -43.1% |
| Sleep min (Apple) | 379.918 | 366.588 | +13.330 | +3.6% |
| TIB min (Apple) | 387.301 | 379.563 | +7.738 | +2.0% |
| Sleep efficiency | 98.067 | 96.530 | +1.537 | +1.6% |
| FL_P95 | 0.106 | 0.090 | +0.016 | +17.5% |
| FL_Avg | 0.0171 | 0.0157 | +0.0015 | +9.7% |
| Min_Above_010 | 20.586 | 15.286 | +5.300 | +34.7% |
| Min_Above_020 | 5.657 | 3.900 | +1.757 | +45.1% |
| Date | AHI | OA | CA | RERA | PAvg | P90 | LAvg | LMax | L90 | FL_P95 | MinAbv010 | MinAbv020 | Deep% | REM% | Awake |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2026-10-03 | 2.33 | 7 | 1 | 2 | 7.21 | 9.30 | 2.77 | 22.8 | 10.8 | 0.08 | 11.3 | 2.5 | 9.63 | 25.46 | 10.54 |
| 2026-10-04 | 1.18 | 6 | 0 | 4 | 7.90 | 10.18 | 1.91 | 28.8 | 4.8 | 0.13 | 25.7 | 7.1 | 7.88 | 27.87 | 1.51 |
| 2026-10-05 | 5.20 | 21 | 0 | 4 | 7.93 | 12.36 | 1.76 | 22.8 | 7.2 | 0.09 | 15.7 | 8.2 | 15.92 | 25.21 | 8.03 |
| 2026-10-06 | 1.85 | 8 | 0 | 1 | 7.19 | 10.24 | 1.04 | 30.0 | 1.2 | 0.08 | 15.1 | 4.6 | 12.12 | 27.14 | 4.01 |
| 2026-10-07 | 1.24 | 7 | 0 | 0 | 7.69 | 10.72 | 1.50 | 34.8 | 3.6 | 0.12 | 23.3 | 4.6 | 18.21 | 23.43 | 6.02 |
| 2026-10-08 | 0.85 | 6 | 0 | 1 | 6.89 | 9.20 | 1.58 | 30.0 | 3.6 | 0.09 | 17.4 | 3.9 | 9.40 | 24.30 | 9.03 |
| 2026-10-09 | 1.49 | 6 | 1 | 5 | 7.38 | 9.88 | 1.34 | 24.0 | 4.8 | 0.08 | 14.0 | 6.7 | 14.02 | 20.09 | 10.54 |
| 2026-10-10 | 1.26 | 5 | 0 | 1 | 7.38 | 10.60 | 1.10 | 18.0 | 4.8 | 0.14 | 32.0 | 9.8 | 10.27 | 18.96 | 4.01 |
| 2026-10-11 | 1.42 | 4 | 0 | 1 | 7.78 | 10.18 | 1.89 | 24.0 | 9.6 | 0.09 | 15.6 | 4.1 | 8.59 | 26.65 | 9.53 |
| 2026-10-12 | 1.41 | 5 | 0 | 0 | 8.20 | 10.74 | 2.18 | 40.8 | 8.4 | 0.14 | 26.7 | 5.9 | 11.04 | 21.17 | 8.53 |
Correlations computed across the full matched set after excluding pre-2026-06-23 low-confidence CPAP nights. The sample has grown by one night (10-12) since the 10-11 report; a few correlation magnitudes shifted modestly.
| Pair | the user r | Interpretation |
|---|---|---|
| AHI vs deep_pct | +0.074 | negligible |
| AHI vs awake_min | +0.063 | negligible |
| OA vs awake_min | +0.125 | weak positive |
| CA vs deep_pct | +0.065 | negligible |
| Leak_Avg vs deep_pct | -0.175 | weak negative |
| Leak_90 vs deep_pct | -0.162 | weak negative |
| Leak_Max vs awake_min | +0.030 | negligible |
| FL_P95 vs awake_min | -0.088 | negligible |
| Min_Above_010 vs deep_pct | +0.005 | negligible |
Bucket comparison (Leak_90 median split, median = 2.40): - Low-leak nights (n=44): mean deep% 15.07, awake 13.99 min - High-leak nights (n=59): mean deep% 14.06, awake 12.20 min - Δ (high – low): deep% -1.01 pp, awake -1.79 min
Tonight's Leak_90 of 8.40 lands deep in the high-leak bucket (3.5× the median split, 3rd-highest Leak_90 in the current window after 10-03's 10.8 and 10-11's 9.6). Tonight's deep% of 11.04 is 3.0 points below the high-leak bucket mean (14.06) — slightly less of a mismatch than yesterday's near-floor 8.59 vs the 13.32 bucket mean. The population-level correlations have shifted modestly with the growing sample: Leak_Avg vs deep_pct is now r = -0.175 (previously -0.080 in the 10-11 report's 87-night subset) and Leak_90 vs deep_pct r = -0.162 (previously -0.063). These shifts reflect that the current cluster of high-leak + lower-deep nights (10-11, 10-12) has strengthened the leak-deep negative signal. MinAbv010 vs deep_pct dropped to +0.005 (previously -0.176), indicating that the previous ranking of "FL as the strongest signal" is now wrong on the full-sample view — leak is now the stronger correlate of deep%, modestly outperforming FL. The |r| magnitudes remain small (≤0.18); leak and FL together still explain only a small fraction of deep-sleep variance at the night-level.