Sleep physiology features from wearable trackers can predict next-day physical fatigue in healthy adults with moderate accuracy.
In a feasibility study of 134 participants (42 healthy, 39 NDD, 53 IMID) across 3,062 nights, machine learning models using sleep tracker data achieved an AUC of 0.75 for predicting next-day physical fatigue in healthy adults, with respiratory rate and REM sleep duration as key predictors. Predictive performance was lower in neurodegenerative disease (AUC=0.62) and immune-mediated inflammatory disease cohorts, indicating limited generalizability to chronic disease populations.
This article is automatically generated from the structured evidence profile behind the claim above. Scores reflect the quality and quantity of available research, not clinical advice.
Sleep physiology features from wearable trackers can predict next-day physical fatigue in healthy adults with moderate accuracy. The current body of evidence comprises 1 study. EvidenceHub rates the overall confidence at 34/100 (low).
The Claim
Sleep physiology features from wearable trackers can predict next-day physical fatigue in healthy adults with moderate accuracy.
This conclusion is most relevant to: 134 participants: 42 healthy adults, 39 with neurodegenerative diseases (NDD), 53 with immune-mediated inflammatory diseases (IMID).
What the Research Shows
The conclusion draws on 1 linked study. Highlights from the cited literature:
- ▸Exploring the feasibility of modeling next-day fatigue and sleepiness using digital sleep tracker data in neurodegenerative and immune-mediated inflammatory diseases. (Frontiers in digital health, 2026) —
How It Works
The proposed biological pathway:
- ▸Sleep trackers record physiological and sleep-architecture features (e.g., respiratory rate, REM duration, REM latency, deep sleep)
- ▸Machine learning models are trained on these features to predict next-day fatigue and sleepiness
- ▸Models use leave-one-subject-out cross-validation to assess predictive performance
- ▸Result: Respiratory rate and REM sleep duration are key predictors of physical fatigue in healthy adults
Who Might Benefit
Evidence fit by population:
- ▸134 participants: 42 healthy adults, 39 with neurodegenerative diseases (NDD), 53 with immune-mediated inflammatory diseases (IMID)
Recommended Dose
N/A
Limitations & Caveats
Important context when interpreting this evidence:
- ▸Outcome binarisation using a global threshold may conflate between-person disease-group differences with within-person symptom variation
- ▸Predictive performance in chronic disease cohorts is limited (AUC ≤ 0.66), suggesting need for larger multimodal studies
- ▸Study is exploratory and not designed for confirmatory hypothesis testing
Frequently Asked Questions
What sleep features best predicted next-day fatigue in healthy adults?▼
Respiratory rate and REM sleep duration were the key drivers of physical fatigue prediction in healthy adults.
How accurate were the models for predicting fatigue in neurodegenerative diseases?▼
The AUC for physical fatigue in NDD was 0.62 under enriched training, with REM latency and deep sleep as key features, indicating limited predictive accuracy.
What types of sleep trackers were used in this study?▼
Three sleep trackers were used: BedSensor, ZKONE, and DREEM 2, with a polysomnography sub-study (n=28) to validate their performance.
Can these findings be applied to clinical practice?▼
No, the findings are exploratory and preliminary. The study underscores the need for larger, multimodal studies to establish disease-specific digital fatigue endpoints before clinical application.
References
- 1.Zhai B, Chen L, Ma X, Pinaud C, Chatterjee M, Kortelainen JM, Rehman RZU, Ahmaniemi T, Avey S, Guan Y, Macrae V, Hinchliffe C, Del Din S, Manyakov NV, Göder R, Romijnders R, Maetzler W, Reilmann R, Aufenberg S, Schubert R, van der Woude CJ, Zhang D, Ng WF. “Exploring the feasibility of modeling next-day fatigue and sleepiness using digital sleep tracker data in neurodegenerative and immune-mediated inflammatory diseases..” Frontiers in digital health, 2026. PMID: 42388291 DOI: 10.3389/fdgth.2026.1752629