Illness or Overreaching? A Wearable-Based 48-Hour Framework Using HRV, Resting HR, Temperature, and Respiratory Rate

If you want the practical answer first: one low-HRV day is not enough to decide whether to train, modify, or rest. The better move is a 48-hour decision window that combines four signals (HRV, resting HR, temperature, respiratory rate), symptom context, and training load history.

That is the operating model SensAI uses: trend-first, uncertainty-aware, and clear on escalation. The goal is simple: avoid missing early illness while also avoiding unnecessary de-training from false alarms.

Why one low-HRV day is not enough evidence

A single HRV drop can reflect many things: hard training, poor sleep, travel, alcohol, menstrual cycle effects, psychological stress, early illness, or plain measurement noise. Treating every low-HRV morning as “you are getting sick” creates avoidable undertraining.

Evidence supports multi-signal interpretation rather than HRV in isolation. In a meta-analysis on HRV-guided training, HRV-guided programs improved vagal-related HRV (SMD 0.50, 95% CI 0.09-0.91) but did not show the same effect for resting HR (SMD 0.04, 95% CI -0.34 to 0.43).¹ Practical implication: HRV is useful, but only with context.

SensAI’s rule: one-day HRV suppression triggers a recheck, not an immediate hard stop.

What overtraining science says about functional overreaching vs maladaptation

Functional overreaching can be part of good training when recovery is adequate; non-functional overreaching and overtraining syndrome are what you want to avoid. Meeusen and colleagues put the core principle clearly: “Successful training not only must involve overload but also must avoid the combination of excessive overload plus inadequate recovery.”²

That sentence is your decision anchor. If load recently increased and non-HRV signals are stable, low HRV may be planned stress. If load is not clearly elevated and multiple illness-linked signals rise together, your probability shifts toward infection or systemic strain.

Illness pattern vs overreaching pattern across 4 wearable signals

The key distinction is clustering. Illness is more likely when autonomic strain appears with thermoregulatory and respiratory disturbance. Overreaching is more likely when autonomic strain appears in a load-linked context without temperature/respiratory drift.

Illness-leaning cluster (HRV down + resting HR up + temperature up + respiratory rate up)

Illness suspicion increases when these changes converge for 24–48 hours:

• HRV below recent baseline band
• Resting HR above baseline trend
• Temperature above personal overnight trend
• Respiratory rate above personal overnight trend

Why this cluster matters:

• In 30,529 participants (3,811 symptomatic), combining wearable data with symptoms improved COVID-positive vs COVID-negative discrimination to AUC 0.80 (IQR 0.73-0.86), versus AUC 0.71 (IQR 0.63-0.79) for symptoms alone.³
• In Fitbit-derived physiology-only modeling, detection reached AUC 0.77 ± 0.018 (sensitivity 0.437 ± 0.037 at 95% specificity), with 2,745 PCR-positive cases among 30,534 tests.⁴
• WHOOP respiratory modeling found low within-person variability (intraindividual SD 0.51 ± 0.20 rpm) and identified 20% of positives in the two days before symptoms and 80% by day 3.⁵
• TemPredict reported AUC 0.819 (95% CI 0.809-0.830); adding continuous dermal temperature raised AUC from 0.770 to 0.819 (+4.9%), with sensitivity 82% and specificity 63%.⁶

As Robert P. Hirten, MD, summarized from Mount Sinai’s wearable work, “subtle changes in a participant’s heart rate variability (HRV) measured by an Apple Watch were able to signal the onset of COVID-19 up to seven days before the individual was diagnosed.”⁷

Overreaching-leaning cluster (HRV down with stable temperature/respiratory rate and load-linked context)

Overreaching is more likely when:

• HRV is down, but temperature and respiratory rate are near baseline
• Resting HR is stable or only mildly elevated
• You can explain the shift with training dose (intensity block, volume spike, poor sleep after hard sessions)
• Symptoms are absent or limited to normal training fatigue

This is where SensAI emphasizes load-linked interpretation. If the timeline fits the training block and non-illness signals are stable, the right move is often to modify, not fully stop.

A 48-hour train/modify/rest decision protocol for athletes

The protocol below is built for morning execution. It reduces emotional overreaction to one metric and gives a clear path when signals disagree.

Hour 0 triage (symptoms + baseline deltas)

At first check (morning):

1. Screen symptoms first.
   • If chest pain, breathing difficulty at rest, confusion, persistent high fever, or other emergency signs are present, do not frame this as a training decision; seek medical care.⁸
2. Check four wearable deltas vs personal baseline.
   • HRV, resting HR, temperature, respiratory rate.
3. Classify initial state.
   • Green (train): 0–1 adverse signals, no meaningful symptoms, clear load context.
   • Amber (modify): 2 adverse signals or uncertain symptom/load story.
   • Red (rest): 3–4 adverse signals and/or symptoms progressing.

SensAI coaching default at Hour 0:

• Green: proceed with planned session but reduce ego pacing.
• Amber: switch to low-to-moderate session and recheck in 24 hours.
• Red: rest or recovery-only; start 24-hour illness watch.

24-hour recheck rules when metrics disagree

Most hard calls happen here. Use these tie-breakers:

• HRV down only, others stable: likely training strain/noise -> train light to moderate.
• HRV + resting HR adverse, temp/resp stable: likely overreaching or sleep stress -> modify, avoid HIIT.
• Temp + respiratory rate rising (with or without HRV): higher illness probability -> rest or very light only.
• Symptoms worsening despite mixed metrics: treat as illness-leaning and de-load.

If disagreement persists, prioritize temperature + respiratory trend + symptom direction over app readiness color.

48-hour escalation thresholds and return-to-training gates

Escalate away from normal training when either pattern appears by 48 hours:

• Persistent illness-leaning cluster across at least 3 signals
• Any meaningful symptom progression

Context from athlete illness literature supports caution. In IOC subgroup meta-analysis (54 studies, n=31,065 athletes), mean acute respiratory illness symptom duration was 7.1 days (95% CI 6.2-8.0), and time loss >1 day occurred in 20.4% (95% CI 15.3-25.4).⁹

Return-to-training gates SensAI uses:

1. Symptoms stable or improving for 24 hours
2. Temperature and respiratory rate trending back toward baseline
3. Resting HR normalizing and HRV no longer declining
4. First session back is reduced-density (volume or intensity cut)

How to map WHOOP, Oura, and Garmin scores to raw physiology

Readiness scores are interfaces. Physiology is the real signal.

Use this translation layer:

• WHOOP Recovery (red/yellow/green) -> map to HRV trend, resting HR trend, and respiratory rate trend.⁵
• Oura Readiness -> map to temperature trend, resting HR trend, HRV trend, and sleep contributors.¹⁰
• Garmin status signals -> map to resting HR pattern, overnight HRV context, and illness-related metric shifts.¹¹

Practical rule from SensAI: do not compare score numbers across platforms. Compare directional changes vs your own baseline, then classify the 4-signal cluster.

If you want a simple check before training: “Do I have autonomic strain only, or autonomic + respiratory/temperature strain together?” That single distinction catches many bad decisions.

Red flags that require medical evaluation (not training decisions)

Stop self-coaching and seek clinical evaluation if any of these are present:

• Shortness of breath at rest or worsening breathing symptoms
• Persistent high fever, chest pain/pressure, confusion, cyanosis, or severe weakness
• Rapidly worsening symptoms despite rest
• Cardiac symptoms during light activity

CDC emergency warning signs are the right safety floor here.⁸ This section exists because wearables support triage, but they are not diagnostic devices.¹¹

The SensAI advantage: cross-device signal fusion with uncertainty-aware coaching

Most athletes do not fail because they lack data. They fail because their data disagree and they still need to decide by 6 a.m.

SensAI solves that with three layers:

1. Cross-device normalization (WHOOP, Oura, Garmin inputs translated to shared physiological features)
2. Uncertainty-aware rules (single-signal drops trigger monitoring; multi-signal convergence triggers decisive action)
3. Action-first coaching output (train/modify/rest now, plus next recheck timing)

Michael Snyder, PhD, captured the opportunity well: “Smartwatches and other wearables make many, many measurements per day—at least 250,000… My lab wants to harness that data and see if we can identify who’s becoming ill as early as possible.”¹²

That is the same philosophy behind SensAI’s coaching engine: extract useful early signal, but avoid overreacting to noise.

Continue with SensAI

• SensAI Home
• About SensAI
• SensAI FAQ
• Contact SensAI
• HRV Across Apple Watch, WHOOP, Oura, and Garmin
• Wildfire Smoke and Training Readiness
• Zone 2 Without 220-Age

Bottom line: in the first 48 hours, your best decision framework is not “low HRV means no training.” It is a structured probability call using HRV, resting HR, temperature, respiratory rate, symptoms, and load context. SensAI helps you make that call early, consistently, and with fewer false positives.

---

Footnotes

1. Manresa-Rocamora A, et al. “HRV-guided training for endurance performance: systematic review and meta-analysis.” International Journal of Environmental Research and Public Health, 2021. https://pubmed.ncbi.nlm.nih.gov/34639599/ ↩

2. Meeusen R, et al. “Prevention, diagnosis and treatment of overtraining syndrome.” Medicine & Science in Sports & Exercise, 2013. https://pubmed.ncbi.nlm.nih.gov/23247672/ ↩

3. Quer G, et al. “Wearable sensor data and self-reported symptoms for COVID-19 detection.” Nature Medicine, 2021. https://pubmed.ncbi.nlm.nih.gov/33122860/ ↩

4. Natarajan A, et al. “Assessment of physiological signs associated with COVID-19 measured using wearable devices.” npj Digital Medicine, 2020. https://www.nature.com/articles/s41746-020-00363-7 ↩

5. Miller DJ, et al. “Analyzing changes in respiratory rate to predict illness using wearable sensors.” PLOS ONE, 2020. https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0243693 ↩ ↩²

6. Smarr BL, et al. “TemPredict: wearable temperature and multimodal data for COVID prediction.” Scientific Reports, 2022. https://www.nature.com/articles/s41598-022-07314-0 ↩

7. Mount Sinai. “Wearable devices can detect COVID-19 symptoms and predict diagnosis.” https://www.mountsinai.org/about/newsroom/2021/mount-sinai-study-finds-wearable-devices-can-detect-covid19-symptoms-and-predict-diagnosis-pr ↩

8. CDC. “COVID-19 Symptoms and Emergency Warning Signs.” https://www.cdc.gov/covid/signs-symptoms/index.html ↩ ↩²

9. Snyders C, et al. “Acute respiratory illness in athletes: IOC subgroup meta-analysis.” British Journal of Sports Medicine, 2022. https://pubmed.ncbi.nlm.nih.gov/34789459/ ↩

10. Oura. “What is the Oura Readiness Score?” https://ouraring.com/blog/readiness-score/ ↩

11. Garmin. “How Getting Sick Might Change Your Heart Metrics.” https://www.garmin.com/en-US/blog/fitness/how-getting-sick-might-change-your-heart-metrics/ ↩ ↩²

12. Stanford Medicine. “Could wearables be the key to detecting infectious disease early?” https://med.stanford.edu/news/all-news/2020/04/wearable-devices-for-predicting-illness-.html ↩