Wearable sensors in sauna research

Wearable sensors in sauna research are body-worn devices used to record physiological signals — most often heart rate, skin temperature, movement and, by estimation, core temperature — during sauna exposure. They appeal to investigators because they are small, familiar to participants and usable where cabled laboratory equipment cannot go, such as between the hot room, showers and cooling areas. Their weakness is that consumer wellness devices were generally developed and validated for rest and exercise in temperate conditions, not for hot, humid, sweaty skin inside an 80–100 °C room, so every signal needs the same sceptical validation as any other research instrument.

What wearables measure directly

Most wrist devices measure heart rate optically. Light-emitting diodes illuminate the skin and a photodetector tracks the pulsing blood volume beneath — photoplethysmography. Done well, this works: a meta-analysis of 44 studies across 15 brands found mean differences against electrocardiography or chest straps of −0.40 beats per minute during sleep, −0.01 at rest and −0.51 during treadmill walking-to-running activity. Chest-strap electrocardiography remains the reference for ambulatory heart-rate monitoring, and sauna studies that need beat-to-beat fidelity generally prefer it.

Error grows where sauna conditions cluster: movement, sweat and variable contact. Upper-body movements produce more errors and signal dropouts than steady locomotion, probably through motion artefacts and changing pressure between device and skin as muscles contract and blood flow shifts. Accuracy falls as intensity rises, and it can collapse for wrist placement during activities the algorithm handles poorly — one cited comparison found 92% accuracy during aerobic exercise but 35% during resistance exercise. Placement matters more than branding: in a 16-participant multi-activity comparison, an upper-arm optical sensor showed a bias of −0.05 bpm with mean absolute percentage error of 1.35% and near-perfect agreement with electrocardiography, while wrist placement on either wrist exceeded 10% error during low-heart-rate tasks such as lying, walking and picking up objects. Sauna bathing resembles the difficult end of this spectrum: participants shift posture, pour water, towel off and move between rooms, all with wet skin.

Two further error sources deserve attention. First, optical signals interact with skin pigmentation and ink: a systematic review of ten studies found four reporting significantly lower heart-rate accuracy in darker-skinned participants, four finding no effect and two with mixed results across devices, with no study favouring darker skin — an unresolved rather than settled question that studies should handle by recruiting diverse samples and reporting skin-tone distributions. Second, sweat, water immersion from cooling plunges and tight or loose bands each change the optical path; a device validated on dry, still wrists has not been validated on wet, moving ones. Sampling rate, averaging windows and missing-data rules must be reported, because a device that records every second and one that reports a smoothed minute value describe different things.

Estimated quantities are not measurements

The most consequential distinction in this literature is between signals a wearable records and quantities its software infers. Estimated core temperature is the leading example. Algorithms such as the US Army's ECTemp use sequential heart-rate observations in a Kalman-filter model — treating heart rate as a noisy observation of core temperature, informed by how the two evolve together during work and heat dissipation — and report accuracy of about −0.03 °C bias with standard deviation 0.32 °C and limits of agreement near ±0.63 °C in their development data. The model was trained on young, fit military personnel, and evaluations outside that population urge caution: in Division I American-football players training in summer heat, agreement between the algorithm's estimate and an ingestible-pill reference reached a concordance of only 0.643, with the authors noting the model needs validation for each commercial device's heart-rate input.

Group-level accuracy can mask individual-level failure. During prolonged real-world walking in warm conditions, the algorithm's group bias of 0.09 °C with limits of ±0.44 °C looked acceptable, yet only 3 of 18 individuals met the 0.1 °C bias criterion and individual correlations ranged from −0.09 to 0.95 — the authors concluded individuals cannot yet rely on their estimated value. Chest-strap-based estimates compared against rectal thermometry across rest, exercise and protective-clothing heat stress differed by −0.1 to +0.3 °C with standard deviations of 0.4 °C, and the evaluators warned that over- or underestimation in particular conditions could permit heat illness if trusted uncritically. For sauna research the implication is direct: an estimated core-temperature trace is a model output conditioned on heart rate, not a temperature recording, and it must be validated against reference thermometry in sauna conditions before any safety or dose claim rests on it. The same applies to energy expenditure, recovery scores and other composite indices, which should be reported as device outputs with their algorithms named.

Heat, humidity and operating limits

A sauna routinely exceeds the environments in which wearables were validated. Multi-sensor core-temperature algorithms reporting good accuracy were demonstrated across ambient temperatures of 13–43 °C — already a wide span, yet still far below hot-room air. Adhesives soften, batteries and displays have rated temperature ceilings, optical windows fog, and charging contacts corrode. Water resistance, where claimed, describes fresh-water immersion at room temperature — typically with time and depth limits — not hours of cyclic dry heat, steam bursts and cold plunges. Researchers should therefore check the model-specific operating range for temperature, humidity and water exposure before fielding any device, position loggers and phones outside the hot room where the protocol allows, and record device failures, dropouts and thermally motivated removals as data rather than silently discarding them. The room's own instrumentation — the temperature sensor, humidity sensor and control unit — and faults in that equipment, covered under sensor failure, are a separate topic from body-worn sensing and should not be conflated with it in methods or discussion. Remote management functions such as safe remote start likewise belong to facility operation, not to physiological measurement.

Motion, sweating and missing data

Sauna protocols generate structured missingness: participants remove wristbands to shower, reposition sensors after towel-drying, and lose contact during pouring or lying prone on benches. Accelerometry from the same device can document posture and movement, turning a nuisance into context — stillness versus fidgeting versus walking to cool down — provided the analyst distinguishes measured movement from the device's own activity classifications. Skin-temperature traces need similar care: evaporative cooling after water throwing, direct radiation near the heater and contact quality all move the trace independently of body state, so thermoregulatory interpretation requires the environmental log alongside. Studies should predefine wear-time rules, artefact thresholds and interpolation limits, and report how much data each rule removed; long silent gaps filled by smoothing can fabricate a reassuring plateau where the sensor had simply lost the wrist.

Ethics and privacy

Continuous wearable recording collects identifiable, high-resolution bodily data, often with timestamps and sometimes location, from participants in states of undress and in facilities where others bathe. Good practice starts with ethics-committee approval and genuine written informed consent that names the signals recorded, who sees them, how long they are kept and whether recordings can be withdrawn — the same standard applied to ingestible-sensor and exercise protocols in this literature. Data minimisation matters: record at the resolution the question needs, strip identifiers before sharing, restrict cloud synchronisation that would export participant data to third-party servers, and consider that heart-rate traces can themselves be identifying. Participants should be able to pause or remove devices without forfeiting participation, and incidental recording of non-participants — in shared hot rooms or changing areas — must be prevented by design rather than apologised for afterwards.

Validation and reporting

A wearable-based sauna study should read like a measurement study, not a product review. Name each device with model and firmware, state what was measured versus estimated, and validate against an appropriate reference in conditions resembling the trial: chest-strap or electrocardiographic heart-rate reference for optical heart rate, reference thermometry for temperature claims, weighed mass balance for sweat-related outputs. Report skin-tone and sex distributions, wear site and tightness, sampling and averaging choices, dropout and artefact rates, and the full environmental record. Link wearable findings with hydration, sweat, blood-pressure and air-quality data where collected, and interpret them within methodological and evidence-quality expectations — including the possibility that the honest result is that a given consumer device does not perform in the sauna. Sauna research benefits from wearables' coverage and comfort; health conclusions still require reference-grade evidence. Studies that also sample the facility environment should keep body-worn and room-level monitoring, such as microbiological monitoring, in their separate methodological lanes.