Reading biomarker combinations together, even when each metric looks fine on its own
Individual wearable metrics are noisy. When several metrics shift together in the same direction, that is easier to put in context than any single number. Here is how to read combinations.
On this page
TL;DR
Individual wearable metrics are noisy - each one shifts for many reasons unrelated to health. When multiple metrics shift together in the same direction across consecutive days, the combination is a qualitatively different signal. Research on illness, overtraining, and recovery has consistently found that multi-metric patterns are more informative than any single number. The key is looking for convergence, not individual excursions. Watching for that convergence by hand is tedious, which is where Sam helps: it reads your HRV, resting heart rate, sleep, and wrist temperature from Apple Health and shows them side by side against your baseline.
Why single metrics mislead
Every wearable metric carries noise. HRV varies based on the sleep stage during which it was measured, the timing of last exercise, recent alcohol consumption, hydration status, and ambient temperature. Resting heart rate varies with how warmly you slept, whether the measurement was taken during a restless versus still period, and residual sympathetic activation from the previous day. Skin temperature varies with room temperature, alcohol intake, and menstrual cycle phase.
This noise means that any single metric on any single night can produce an unusual reading with no meaningful cause. An HRV that looks 15% below your recent average might reflect a stressful day, a glass of wine, a slightly disturbed night of sleep, or nothing in particular - as often as it reflects genuine physiological disruption.
Now consider what has to happen for multiple metrics to all shift in the same direction on the same nights. For HRV to fall, resting heart rate to rise, skin temperature to elevate, and sleep to fragment simultaneously, there would need to be multiple different random factors all pointing in the same direction at once. That is considerably less likely than a single random fluctuation. When it happens repeatedly across two, three, or four consecutive nights, the probability that it represents random noise decreases substantially.
This is the core logic of multi-metric interpretation: convergence across independent data streams is a more specific signal than any individual stream's movement.
What the research shows about combined signals
Research on wearable devices in illness detection has repeatedly found that multi-signal approaches outperform single-metric approaches.
During the COVID-19 pandemic, multiple research groups studied whether consumer wearables could detect physiological changes associated with infection. A study by Mayer and colleagues published in Cell Reports Medicine (PMC 9017023) analysed wearable data from participants who contracted COVID-19 and found that multiple physiological features - resting heart rate, heart rate circadian variation, and heart rate autocorrelation - changed in recognisable patterns both before and after symptom onset. No single feature was reliable alone; the combination created a more interpretable signal.
A systematic review in The Lancet Digital Health (PMC 9020803) examined 14 studies on wearable sensor performance in SARS-CoV-2 detection and found that combining skin temperature, resting heart rate, and respiratory rate improved detection sensitivity compared to using heart rate alone.
The most-cited individual-level finding comes from Mishra et al. (Stanford, Snyder lab), published in Nature Biomedical Engineering in 2020. They identified 32 COVID-19 cases within a cohort of nearly 5,300 consumer-smartwatch users, and 26 of the 32 (81%) showed alterations in heart rate, daily steps, or time asleep - with 22 of 25 cases that had symptom data detected at or before symptom onset, and four cases detected at least nine days early. The key phrase is "above individual baseline" - not above a population average, and not from an isolated reading, but from a sustained elevation.
None of this research enables wearables to diagnose illness. What it demonstrates is that physiological disruption - of whatever origin - tends to produce correlated changes across multiple autonomic and metabolic readouts simultaneously, and that tracking those combinations makes the pattern more recognisable.
The common convergent patterns
Based on physiology and the research above, a few multi-metric combinations are worth recognising:
HRV down + RHR up (two or more consecutive nights)
This combination reflects a shift in autonomic balance toward sympathetic dominance: the nervous system is more active, the heart is beating faster at rest, and vagal tone is reduced. This pattern is consistent with accumulated physiological stress, early immune response, inadequate recovery after hard training, or sustained psychological stress. It is not specific to any single cause, but the combination is more meaningful than either metric alone. The physiology behind why HRV and resting heart rate move in opposite directions - and what drives each - is covered in HRV and resting heart rate: how they relate.
A note on the timeframe: two consecutive nights is short on purpose. The point of multi-metric convergence is that it tightens the signal - when several independent metrics all shift the same way at once, you don't need a long window to separate signal from noise. Single-metric patterns need longer windows: a 5-7-day run of suppressed HRV is the relevant marker for training overload (overtraining when you're not an athlete), and a 2-4-week trend is what's calibrated for chronic occupational stress (burnout and Apple Watch). Same metric, different windows for different mechanisms.
HRV down + RHR up + sleep quality disrupted
Adding sleep disruption to the HRV/RHR combination strengthens the pattern. The body is clearly not recovering effectively: autonomic balance is off, resting state is elevated, and sleep is fragmented. This combination is commonly seen in the days preceding or during an acute illness, in overreach following a very high training load, and in sustained periods of high life stress.
HRV down + RHR up + skin temperature elevated + sleep disrupted
The full four-metric convergence is the strongest cluster. When an immune response is active, inflammatory cytokines signal the hypothalamus, autonomic balance shifts, peripheral perfusion changes, and sleep architecture is disturbed - all simultaneously. This cluster, sustained over two or more nights, warrants serious attention to how you feel and how your body is responding, and is worth prioritising rest and recovery regardless of the specific cause.
Step count down without other metric shifts
A drop in step count while HRV, RHR, and sleep remain near baseline is a different pattern: almost certainly behavioural rather than physiological. A busy day, bad weather, a work deadline, or simply staying home can produce it. It does not carry the physiological implication of the combinations above.
RHR elevated alone, single night
A single elevated resting heart rate reading without HRV suppression or other shifts is most often a mundane single-night event: a warm room, residual dehydration, caffeine timing, a restless period of sleep during measurement. It is not the same pattern as sustained multi-metric convergence.
How to use this in practice
The practical implication is simpler than it might seem: rather than reacting to individual daily metric readings, develop the habit of looking at whether metrics are moving in the same direction across multiple days.
A few things that help:
Look at three-day and seven-day windows. Glancing at this morning's single reading is less useful than scanning whether the last three nights have all shown HRV below your rolling average and RHR above it. That pattern is a signal; a single reading is usually not.
Notice convergence before acting on any single number. If HRV is down today but resting heart rate is normal, sleep was fine, and you feel well, the most likely explanation is noise. If all four metrics point the same direction for three days and you feel more fatigued than usual, that convergence is meaningful.
Treat the convergent signal as an invitation to rest, not a diagnosis. The appropriate response to a multi-metric physiological stress pattern - whatever its cause - is typically the same: prioritise sleep, reduce training intensity, manage workload, and monitor whether the pattern resolves. This is appropriate regardless of whether the cause is accumulating training stress, early illness, sustained work pressure, or something else.
Let subjective experience confirm or question the pattern. How you feel is itself data. Multi-metric convergence that correlates with subjective fatigue, reduced motivation, or generally feeling flat is a stronger signal than the same pattern in someone who feels great and is performing well.
See HRV, resting heart rate and sleep together in Sam
You look at the signals from one night side by side and compare each with your own recent weeks.
- Instead of Apple Health alone: Sam reads heart rate, resting heart rate, HRV, respiratory rate and wrist temperature during sleep from Apple Health and lists them in one place together with sleep and activity.
- Instead of screenshots: each metric shows how the recent days compare with your own average of the past weeks, so every signal gets the same comparison.
- Instead of memory at the appointment: the monthly report you create yourself takes sleep, activity and your logged symptom days to the appointment as a PDF.
What a combination of changes means is something to discuss with your doctor.
Try Sam for freeSam is a wellness companion, not a medical device. The app does not diagnose, treat, cure or prevent any condition or health state. Always consult qualified professionals for medical decisions.
Related reading
Sources
- Mayer C, Tyler J, Fang Y, Flora C, Frank E, Tewari M, Choi SW, Sen S, Forger DB. Consumer-grade wearables identify changes in multiple physiological systems during COVID-19 disease progression. Cell Rep Med. 2022;3(4):100601.
- Perez MV, et al. The performance of wearable sensors in the detection of SARS-CoV-2 infection: a systematic review. Lancet Digit Health. 2022.
- Mishra T, Wang M, Metwally AA, Bogu GK, Brooks AW, Bahmani A, et al. Pre-symptomatic detection of COVID-19 from smartwatch data. Nat Biomed Eng. 2020;4(12):1208-1220.
- Radin JM, Wineinger NE, Topol EJ, Steinhubl SR. Harnessing wearable device data to improve state-level real-time surveillance of influenza-like illness in the USA: a population-based study. Lancet Digit Health. 2020;2(2):e85-e93.
- Rijo-Ferreira F, Takahashi JS. A tangled threesome: circadian rhythm, body temperature variations, and the immune system. Biology. 2021;10(1):65.
More on recording and taking it along
Frequently Asked Questions
Why do multiple wearable metrics shift together during illness?+
When the immune system mounts a response to an infection, it triggers a broad physiological reaction: heart rate rises, autonomic balance shifts toward sympathetic dominance (suppressing HRV), sleep architecture is disrupted, and peripheral temperature may rise. These are not independent signals - they are different readouts of the same underlying physiological state.
What is the most reliable multi-metric pattern to notice on a wearable?+
The combination of declining HRV alongside rising resting heart rate, sustained over two or more consecutive nights and accompanied by disrupted sleep quality, is one of the more consistent patterns associated with physiological stress, early illness, or insufficient recovery. Adding a skin temperature elevation above baseline strengthens the signal further.
Why can a single metric look normal while something is still off?+
Individual metrics are noisy - each one fluctuates for many reasons that have nothing to do with health. HRV varies with sleep stage at measurement; resting heart rate varies with hydration and temperature; skin temperature varies with room conditions. A single metric at an unusual value is often noise. Multiple metrics shifting together is a more specific signal because the same random factors are unlikely to move all of them simultaneously.
Can wearables detect illness before symptoms appear?+
Not reliably. In studies during the COVID-19 pandemic, resting heart rate elevations and HRV suppressions appeared a median of two to three days before symptom onset in some participants. However, sensitivity and specificity varied widely and these findings are not sufficient for clinical detection.
What combination of metrics should I track daily?+
For a practical multi-metric overview: resting heart rate (overnight), HRV (overnight), sleep duration and quality, and - where available - skin temperature deviation. These four metrics together, viewed as a short trend rather than a single night, give a reasonable picture of physiological recovery and stress load.
Is there a combination that specifically signals overtraining?+
Overtraining in the conventional sense is a clinical condition requiring formal assessment, but patterns associated with accumulated training stress often include persistently suppressed HRV alongside elevated resting heart rate and declining performance or motivation. These patterns tend to develop over weeks rather than appearing overnight. A single bad night after a hard workout is normal; a ten-day trend of declining HRV with rising RHR despite rest days is a different signal.
What if only one metric shifts but I feel fine?+
An isolated shift in a single metric - particularly a mild one, on a single night - is most likely noise, measurement variability, or a mundane explanation like alcohol, late exercise, or a warmer room. How you feel subjectively is itself a data point. One metric shifting while everything else holds and you feel well is rarely a meaningful signal.