GvHD Prediction Model

Oura Ring, 2026-01-08 to 2026-10-06 (272 days) | State model: HMM | Ruxolitinib started 2026-03-16 | HEV diagnosed 2026-03-18

Generated 2026-10-07 06:22 · Post-HSCT Patient · Data Jan 08, 2026 to Oct 06, 2026 · Day 205 post-treatment

Peak GVHD Composite

Critical

73.2

on 2026-03-02

HMM Feb 9 State

Flare

Active Flare

GaussianHMM (hmmlearn)

RED Alerts in ±3d Window

Detected

7/139

132 outside ±3d event window | retrospective only

Combined GVHD + BOS

Normal

39.2

LOW

Clinical Context: Chronic GVHD affecting skin, liver, mouth (NIH 2014: moderate). Known acute decompensation on Feb 9, 2026 (validation target). HMM classified Feb 9 as Active Flare, while RED alerts were mostly off-window (132/139 outside ±3d), so treat this as a retrospective state-classification signal. Top ranked features in this run: Sleep Heart Rate, Lowest HR, Readiness Score. Temperature deviation contributes inflammatory context but is not the leading feature.

REPORT SECTION

1. Temperature Fluctuation Analysis

Temperature deviation from Oura measures deviations from personal baseline body temperature. GVHD causes micro-inflammatory signatures visible as increased night-to-night variability. CUSUM change-point detection identifies regime shifts in temperature dynamics.

Regime changes detected: 2026-01-31, 2026-02-11, 2026-03-02, 2026-04-13, 2026-04-22, 2026-04-28, 2026-05-05, 2026-05-21, 2026-05-29, 2026-06-01, 2026-06-08, 2026-06-11, 2026-07-10, 2026-07-11, 2026-08-01, 2026-08-16, 2026-08-17, 2026-08-24, 2026-09-18, 2026-10-02

REPORT SECTION

2. Multi-Stream GVHD Composite Score

Six biometric streams weighted by clinical relevance: temperature (25%), HRV (20%), SpO2 (15%), sleep fragmentation (15%), resting HR (15%), activity (10%). Each component z-scored against first 14 days baseline, normalized to 0-100 (higher = more GVHD-like).

REPORT SECTION

3. HMM State Model

4-state Gaussian HMM (hmmlearn) fitted to composite score + temperature deviation + HRV (3 features). rSLDS was unavailable in this run. Reason: rSLDS path unavailable in this run.

HMM State Distribution

State	Days	Percentage
Remission	79	29.0%
Pre-flare	52	19.1%
Active Flare	89	32.7%
Recovery	52	19.1%

REPORT SECTION

4. Retrospective Alert Burden

YELLOW alert: P(pre-flare) > 0.3 for 3+ consecutive days.
RED alert: P(pre-flare) > 0.5 OR P(active flare) > 0.2.
Retrospective validation against Feb 9, 2026 acute decompensation: 7/139 RED alerts fell within ±3d of the event and 132 occurred outside that window.

Alert History (139 total)

Date	Level	P(Pre-flare)	P(Flare)	Composite
2026-01-08	RED	0.0	1.0	43.1
2026-01-09	RED	0.0	1.0	65.5
2026-01-10	RED	0.0	1.0	59.5
2026-01-11	RED	0.0	1.0	59.0
2026-01-12	RED	0.0	1.0	42.3
2026-01-13	RED	0.0	1.0	66.7
2026-01-14	RED	0.0	1.0	40.2
2026-01-15	RED	0.0	1.0	52.2
2026-01-16	RED	0.0	1.0	44.7
2026-01-17	RED	0.0	1.0	44.5
2026-01-18	RED	0.0	1.0	55.4
2026-01-19	RED	0.0	1.0	22.3
2026-01-20	RED	0.0	1.0	56.5
2026-01-21	RED	0.0	1.0	54.6
2026-01-22	RED	0.0	1.0	52.3
2026-01-23	RED	0.0	1.0	49.8
2026-01-24	RED	0.0	1.0	39.6
2026-01-25	RED	0.0	1.0	40.6
2026-01-26	RED	0.0	1.0	28.8
2026-01-27	RED	0.0	1.0	36.8
2026-01-28	RED	0.0	1.0	34.5
2026-01-29	RED	0.0	1.0	62.7
2026-01-30	RED	0.0	1.0	26.9
2026-01-31	RED	0.0	1.0	68.2
2026-02-01	RED	0.0	1.0	48.8
2026-02-02	RED	0.0	1.0	28.1
2026-02-03	RED	0.0	1.0	37.0
2026-02-04	RED	0.0	1.0	49.6
2026-02-05	RED	0.0	1.0	41.6
2026-02-06	RED	0.0	1.0	34.6
2026-02-07	RED	0.0	1.0	35.0
2026-02-08	RED	0.0	1.0	38.8
2026-02-09	RED	0.0	1.0	69.8
2026-02-10	RED	0.0	1.0	35.0
2026-02-11	RED	0.0	1.0	47.0
2026-02-12	RED	0.0	1.0	34.2
2026-02-13	RED	0.0	1.0	34.8
2026-02-14	RED	0.0	1.0	38.6
2026-02-15	RED	0.0	1.0	48.8
2026-02-16	RED	0.0	1.0	49.7
2026-02-17	RED	0.0	1.0	43.7
2026-02-18	RED	0.0	1.0	47.1
2026-02-19	RED	0.0	1.0	51.3
2026-02-20	RED	0.0	1.0	43.0
2026-02-21	RED	0.0	1.0	35.1
2026-02-22	RED	0.0	1.0	34.5
2026-02-23	RED	0.0	1.0	40.7
2026-02-24	RED	0.0	1.0	29.1
2026-02-25	RED	0.0	1.0	23.4
2026-02-26	RED	0.0	1.0	33.9
2026-02-27	RED	0.0	1.0	45.3
2026-02-28	RED	0.0	1.0	43.1
2026-03-01	RED	0.0	1.0	35.6
2026-03-02	RED	0.0	1.0	73.2
2026-03-03	RED	0.999	0.001	44.8
2026-03-05	RED	0.0	0.938	17.5
2026-03-06	RED	0.0	0.983	25.6
2026-03-07	RED	0.0	0.999	21.3
2026-03-08	RED	0.0	1.0	33.5
2026-03-09	RED	0.0	1.0	31.1
2026-03-10	RED	0.0	1.0	23.2
2026-03-11	RED	0.0	1.0	40.6
2026-03-12	RED	0.0	1.0	42.5
2026-03-13	RED	0.0	1.0	42.8
2026-03-14	RED	0.0	1.0	49.6
2026-03-15	RED	0.0	1.0	30.8
2026-03-16	RED	0.0	1.0	29.3
2026-03-17	RED	0.0	1.0	43.9
2026-03-18	RED	0.0	1.0	46.6
2026-03-19	RED	0.0	1.0	46.0
2026-03-20	RED	0.0	1.0	25.8
2026-03-21	RED	0.0	1.0	31.9
2026-03-22	RED	0.016	0.984	50.8
2026-03-23	RED	0.0	0.984	23.8
2026-03-24	RED	0.0	1.0	34.2
2026-03-25	RED	0.0	1.0	42.3
2026-03-26	RED	0.0	1.0	41.8
2026-03-27	RED	0.0	1.0	30.9
2026-03-28	RED	0.0	1.0	38.7
2026-03-29	RED	0.0	1.0	30.6
2026-03-30	RED	0.0	1.0	39.8
2026-03-31	RED	0.0	1.0	38.6
2026-04-01	RED	0.0	1.0	22.8
2026-04-02	RED	0.0	1.0	49.5
2026-04-03	RED	0.0	1.0	30.6
2026-04-04	RED	0.0	1.0	33.0
2026-04-05	RED	0.0	1.0	39.1
2026-04-06	RED	0.0	1.0	64.9
2026-04-07	RED	0.0	1.0	44.7
2026-04-08	RED	0.005	0.995	66.2
2026-04-09	RED	1.0	0.0	35.0
2026-04-13	RED	1.0	0.0	40.1
2026-04-16	RED	0.763	0.0	27.9
2026-04-24	RED	1.0	0.0	29.0
2026-04-26	RED	0.868	0.0	27.0
2026-04-27	RED	0.979	0.0	32.4
2026-04-28	RED	0.996	0.0	41.0
2026-04-29	RED	1.0	0.0	35.9
2026-05-01	RED	0.897	0.0	36.7
2026-05-10	RED	0.963	0.0	48.0
2026-05-19	RED	1.0	0.0	36.6
2026-05-21	RED	0.997	0.0	37.0
2026-05-22	RED	0.992	0.0	40.5
2026-05-26	RED	1.0	0.0	51.5
2026-05-28	RED	1.0	0.0	33.9
2026-05-29	RED	1.0	0.0	41.5
2026-05-31	RED	1.0	0.0	36.9
2026-06-01	RED	1.0	0.0	45.2
2026-06-06	RED	0.933	0.0	33.7
2026-06-10	RED	0.998	0.0	47.3
2026-06-11	RED	1.0	0.0	47.0
2026-06-12	RED	1.0	0.0	50.9
2026-06-20	RED	1.0	0.0	50.4
2026-06-21	RED	0.55	0.0	33.2
2026-06-22	RED	0.563	0.0	41.1
2026-06-23	RED	0.997	0.0	50.1
2026-07-01	RED	0.885	0.0	39.6
2026-07-11	RED	0.998	0.002	69.7
2026-07-12	RED	1.0	0.0	49.1
2026-07-18	RED	1.0	0.0	46.4
2026-07-26	RED	0.825	0.0	36.4
2026-08-02	RED	0.964	0.005	49.9
2026-08-04	RED	0.998	0.0	38.4
2026-08-13	RED	0.959	0.0	40.0
2026-08-16	RED	1.0	0.0	51.0
2026-08-17	RED	1.0	0.0	32.5
2026-08-23	RED	1.0	0.0	55.2
2026-08-24	RED	1.0	0.0	34.1
2026-08-26	RED	0.986	0.0	42.9
2026-09-07	RED	0.812	0.0	41.1
2026-09-08	RED	0.534	0.0	41.5
2026-09-13	RED	1.0	0.0	47.8
2026-09-15	RED	0.554	0.0	34.9
2026-09-18	RED	1.0	0.0	44.3
2026-09-21	RED	0.893	0.0	37.9
2026-09-23	RED	0.542	0.0	38.0
2026-09-25	RED	1.0	0.0	49.9
2026-09-27	RED	0.929	0.0	43.2
2026-10-02	RED	1.0	0.0	43.4

REPORT SECTION

5. Predictive Feature Importance

Features ranked by combined score: point-biserial correlation (30%), mutual information (30%), composite correlation (20%), Cohen's d effect size (20%). Binary target: composite score > 65th percentile.

Top Predictive Features for GVHD State

Feature	Importance	Correlation	Mutual Info	Cohen's d
Sleep Heart Rate	0.421	+0.498	0.129	+1.14
Lowest HR	0.394	+0.474	0.108	+1.09
Readiness Score	0.382	-0.456	0.105	-1.03
HRV Median (RMSSD)	0.285	-0.342	0.082	-0.76
Sleep Efficiency	0.283	-0.374	0.062	-0.79
Sleep Fragmentation	0.266	+0.329	0.069	+0.71
HRV Variability	0.251	-0.294	0.088	-0.64

REPORT SECTION

6. BOS Risk Integration

Bronchiolitis obliterans syndrome (BOS) is the pulmonary manifestation of chronic GVHD. BOS risk score (39.3/MODERATE) loaded from SpO2/BOS analysis and integrated with systemic GVHD composite. SpO2 trend analysis provides supplementary pulmonary assessment.

BOS Risk Score: 39.3 (MODERATE)
Combined Risk: 39.2 (LOW)
SpO2 Mean: 95.62%
SpO2 Trend: -0.0084 %/day

Disclaimer: This analysis is for research purposes only and should not be used as a sole basis for clinical decisions. N=1 retrospective case study - all detection metrics are descriptive, not inferential. Sensitivity/specificity cannot be computed from a single patient. Validation requires an external multi-patient cohort. Temperature deviation from consumer wearables has limited precision compared to clinical thermometry. All clinical decisions should be made in consultation with the treating hematologist.