A worn sensor network streaming VOC, particulate, UV, noise and thermal data into a personal exposome risk profile
Epidemiologist Christopher Wild coined the term "exposome" in a 2005 Cancer Epidemiology, Biomarkers & Prevention editorial, arguing that genome-scale mapping of disease risk was incomplete without an equally systematic accounting of lifetime environmental exposure. Two decades later, low-cost MEMS and optical sensors have made continuous, individual-level exposure monitoring practical outside the research lab — the foundation of the personal exposome wearable.
A modern exposome wristband or badge packs several independent transducers into a device the size of a smartwatch:
• Optical particle counter (e.g. Plantower PMS5003): a laser diode scatters light off individual particles drawn through a small fan-driven chamber; scattering intensity and pulse count are binned into particle-size fractions (PM1.0, PM2.5, PM10) via Mie scattering theory. Detection range ~0.3–10 µm, resolution ~1 µg/m³.
• Metal-oxide (MOx) VOC array (e.g. Bosch BME680, SGX MiCS): a heated tin-oxide or tungsten-oxide film changes resistance when reducing gases (ethanol, formaldehyde, benzene fragments) adsorb to its surface. Output is reported as a unitless IAQ (indoor air quality) index or estimated total VOC (eTVOC) in ppb — a proxy rather than a specific-compound quantification.
• UV photodiode: a GaP or AlGaN photodiode with an erythemal-weighting filter approximates the CIE erythemal action spectrum, producing a UV Index reading and integrating exposure into Standard Erythemal Dose (SED = 100 J/m² of erythemally-effective UV).
• MEMS microphone dosimeter: samples sound pressure level in dBA, integrating over time for noise dose relative to OSHA's 85 dBA 8-hour time-weighted average (TWA) action level.
• Skin-contact thermistor: NTC thermistor measures skin temperature as a proxy for thermal stress and, combined with ambient temperature, heat-index style exposure.
• GPS/GNSS receiver: provides location fixes at 30–60 s intervals, later matched to a geofenced microenvironment classifier (home / transit / work / outdoor).
A single PMS5003-class optical counter costs roughly $10–15 in volume yet approaches ±10 µg/m³ agreement with reference beta-attenuation monitors (BAM-1020) under most ambient conditions — the price collapse that made population-scale personal monitoring feasible.
Continuous 1 Hz sampling across six channels would exhaust a coin-cell or small LiPo battery within hours, so exposome wearables duty-cycle aggressively:
• PM optical counters draw the most current (the fan motor draws 60–100 mA); typical duty cycle is a 10–30 s burst every 1–5 minutes, interpolated between bursts. • MOx VOC sensors require a heater element (~150–400°C) that dominates power draw; low-power designs pulse-heat the element and read resistance during a brief cool-down window. • UV photodiodes and thermistors draw microamps and can sample continuously at low rates. • GPS is the second-largest power draw and is often duty-cycled to one fix per 30–60 s, with dead-reckoning via accelerometer between fixes.
The tradeoff between temporal resolution and battery life is central to wearable exposome design: RTI International and EPA personal-exposure panel studies (e.g. the Detroit Exposure and Aerosol Research Study, DEARS) historically relied on bulkier, actively-pumped monitors worn for 24–48 hour windows; today's passive optical/MOx wearables trade some accuracy for weeks of continuous, low-burden data collection.
Raw sensor output is not directly a regulatory-grade concentration. Each channel requires calibration against reference instrumentation:
• PM2.5: co-location against a Federal Equivalent Method (FEM) monitor or research-grade nephelometer (e.g. TSI DustTrak) yields a correction factor; humidity strongly biases optical scattering (hygroscopic growth), so most algorithms apply a relative-humidity correction (e.g. the EPA correction equation developed for PurpleAir sensors). • VOC: MOx resistance is converted to an index via manufacturer lookup curves; without a reference gas chromatograph, absolute ppb values for specific compounds (formaldehyde, benzene) cannot be recovered — only a composite reactivity signal. • UV: photodiode counts are converted to UV Index using a factory erythemal calibration traceable to a spectroradiometer.
This calibration gap is why Stage 3 (biomarker correlation) matters: internal biological measurements provide an independent check on what the external sensors are actually reporting.
People do not experience one exposure level per day — they move through a sequence of microenvironments (home, commute, workplace, outdoors) each with a distinct chemical and physical signature. Time-activity diaries, historically self-reported and error-prone, are now automatically reconstructed by fusing GPS traces with continuous sensor streams, revealing exposure peaks invisible to any fixed-site ambient monitor.
A single 24-hour average exposure value obscures the acute peaks that often drive health risk. The classic National Human Activity Pattern Survey (NHAPS, Klepeis et al. 2001) found Americans spend roughly 87% of their time indoors and about 6% in enclosed transit — yet these brief transit windows can contribute a disproportionate share of daily inhaled pollutant dose because in-cabin concentrations during highway commuting frequently exceed roadside ambient levels by 2–5× due to self-pollution from surrounding traffic and cabin air recirculation.
Exposome wearables reconstruct a continuous microenvironment sequence by fusing:
• GPS/GNSS fixes matched against geofenced polygons (home, workplace) or transit-mode classifiers (speed/acceleration signatures distinguishing walking, cycling, driving, rail) • Sensor signal shape — e.g. a rapid PM2.5 spike coincident with GPS-inferred kitchen dwell time strongly suggests a cooking event, not ambient infiltration • Time-of-day priors combined with a hidden Markov model over microenvironment states, smoothing noisy instantaneous classifications into coherent dwell periods
Each microenvironment imprints a characteristic multi-channel fingerprint on the sensor stream:
• Home (night/morning): low PM2.5 baseline (5–15 µg/m³) punctuated by sharp cooking-related VOC and PM spikes (gas-stove NO2/PM co-emission, frying aerosol); low noise (35–50 dBA); minimal UV. • Commute (car/transit/bike): PM2.5 spikes to 20–60 µg/m³ in traffic-dense corridors from vehicle exhaust and brake/tire wear resuspension; elevated cabin VOCs from off-gassing interior materials in new/hot vehicles; noise 65–85 dBA (higher for motorcycle/open-window cycling); UV exposure highest for cyclists/pedestrians. • Workplace/office: generally the cleanest microenvironment for PM (mechanically filtered HVAC, often <10 µg/m³) but VOC index can rise from off-gassing furnishings, printers/copiers (ozone, VOCs), and CO2 buildup in poorly ventilated conference rooms; noise moderate (45–65 dBA) with periodic spikes. • Outdoor/park: PM2.5 tracks regional ambient background, modulated by proximity to roadways; UV dose accumulates rapidly (peak UV Index 6–11 midday); noise variable but often lower than commute.
The HEALS (Health and Environment-wide Associations based on Large population Surveys) EU exposome project and the earlier RIOPA (Relationships of Indoor, Outdoor and Personal Air) study both found that personal PM2.5 exposure correlated only weakly (r ≈ 0.3–0.5) with same-day fixed-site ambient monitors — direct evidence that microenvironment-resolved personal monitoring captures dose that stationary regulatory networks miss.
Once each timestamp is labeled with a microenvironment and an instantaneous concentration, cumulative daily dose is computed as a time-weighted integral:
Dose = Σᵢ (Cᵢ × Δtᵢ) across all microenvironment dwell intervals i
This simple summation already reveals actionable structure: a 40-minute high-traffic commute at 45 µg/m³ PM2.5 can contribute nearly as much integrated dose as 10 hours at a 3 µg/m³ office baseline. Personal exposure panel studies run by RTI International and the EPA (e.g. the Relationships of Indoor, Outdoor, and Personal Air, and Detroit Exposure and Aerosol Research studies) established this microenvironment-weighted-dose framework using bulky pumped samplers in the 1990s–2000s; wearable optical/MOx sensors now let the same framework run continuously, in free-living conditions, at population scale.
A wearable sensor measures what surrounds the body; a biomarker measures what the body actually absorbed. Reconciling the two — external exposure estimated from PM2.5, VOC and other sensor channels against internal measurements like urinary cotinine, urinary 1-hydroxypyrene, and exhaled breath VOCs — is the validation step that turns a sensor reading into a defensible dose estimate, and is central to how the NHANES biomonitoring program and the exposome research field build confidence in wearable-derived exposure data.
Biomonitoring measures a chemical, its metabolite, or a biological response product in blood, urine, or exhaled breath — a direct readout of absorbed (not just ambient) dose:
• Urinary cotinine: the primary nicotine metabolite, the gold-standard biomarker for tobacco smoke exposure — both active smoking and secondhand exposure. CDC/NHANES bands: <1 ng/mL typically indicates no exposure, 1–30 ng/mL secondhand smoke exposure, >50 ng/mL consistent with active smoking, though cutoffs vary by lab and population.
• 1-hydroxypyrene (1-OHP): the major urinary metabolite of pyrene, used as a biomarker for exposure to polycyclic aromatic hydrocarbons (PAHs) — a chemical class released by combustion (traffic exhaust, grilled/charred food, wood smoke, tobacco). 1-OHP integrates PAH exposure across all routes (inhalation, dermal, dietary) over the prior 24–48 hours, reflecting pyrene's short biological half-life.
• Exhaled breath VOCs: real-time breath analyzers (e.g. selected-ion flow-tube or PTR mass spectrometry in research settings) capture volatile compounds cleared via the lungs — including benzene, toluene, and metabolic byproducts — offering a near-instantaneous window into recent inhalation exposure and, increasingly, endogenous metabolic state.
• Blood/urine metal panels: NHANES tracks urinary and blood cadmium, lead, mercury and arsenic as biomarkers of cumulative heavy-metal exposure from diet, occupational, and environmental sources.
A raw biomarker concentration is only interpretable against a reference framework. Biomonitoring Equivalents (BE values), developed collaboratively by toxicologists and risk assessors, translate existing exposure guidance values (like an EPA Reference Dose) into the corresponding expected biomarker concentration — allowing a measured urinary or blood level to be compared directly against a health-based benchmark rather than only against population percentiles.
NHANES (the National Health and Nutrition Examination Survey), run continuously by the CDC since the 1960s and biomonitoring specific analytes since 1999, provides the population reference distributions (e.g. 50th, 95th percentile) against which an individual's wearable-linked biomarker panel is typically benchmarked. A measured 1-OHP or cotinine value can then be expressed as a percentile relative to the broader US population, contextualizing whether an individual's exposure is unusually elevated.
Because cotinine has a biological half-life of ~16–20 hours and 1-OHP is cleared within 1–2 days, both biomarkers act as short integrating windows — ideal for correlating against a wearable's trailing 24–48 hour sensor-derived dose estimate, rather than a single instantaneous reading.
Validating a wearable's exposure estimate against biomarkers follows a structured panel-study design, the same approach used by EPA- and RTI-affiliated personal exposure panel studies:
1. Participants wear the sensor array continuously for 1–2 weeks while providing periodic urine/blood/breath samples (e.g. first-morning void, matched to the prior day's integrated sensor dose). 2. Sensor-derived cumulative dose (Stage 2's time-weighted integral) is regressed against the paired biomarker concentration. 3. A Pearson or Spearman correlation coefficient (typically r ≈ 0.6–0.8 for PM/PAH-linked biomarkers in published panel studies) quantifies how well the external sensor signal predicts internal absorbed dose. 4. Systematic offsets reveal calibration or route-of-exposure gaps — e.g. dietary PAH intake (grilled meat) elevates 1-OHP independent of any inhalation sensor reading, a reminder that wearables capture only the inhalation/dermal exposure route, not diet.
This correlation step is what elevates a consumer air-quality gadget into a genuine exposome research instrument: it quantifies, rather than assumes, how faithfully ambient sensing tracks what actually enters the bloodstream.
No single exposure channel tells the whole story. A composite exposome risk score integrates dose-normalized values across PM2.5, VOC/formaldehyde, noise, UV and thermal stress axes — each benchmarked against its own regulatory or health-based reference limit — into one interpretable number, mirroring how multi-exposure risk indices are constructed in occupational and environmental health practice.
Each channel in the composite score is normalized against an established occupational or public-health benchmark:
• PM2.5: WHO's 2021 Air Quality Guideline set the annual mean target at 5 µg/m³ (tightened from the prior 10 µg/m³ guideline) and a 24-hour guideline of 15 µg/m³; the US EPA's National Ambient Air Quality Standard (NAAQS) sets the annual PM2.5 standard at 9 µg/m³ (revised 2024) and 35 µg/m³ for 24-hour average.
• VOC/formaldehyde: no single universal exposure limit exists for the composite MOx VOC index, but formaldehyde — a common indoor VOC contributor — carries a WHO indoor air guideline of 0.1 mg/m³ (30-minute average) and an OSHA permissible exposure limit (PEL) of 0.75 ppm (8-hr TWA) in occupational settings.
• Noise: OSHA's Permissible Exposure Limit is 90 dBA as an 8-hour TWA (with a 5 dB exchange rate), while the OSHA/NIOSH-recommended action level is 85 dBA (NIOSH uses a more conservative 3 dB exchange rate, meaning allowable duration halves for every 3 dB increase above 85 dBA).
• UV: exposure is tracked in Standard Erythemal Dose (SED) units, where 1 SED = 100 J/m² of erythemally-weighted UV — roughly the threshold at which fair skin begins to redden (related to the Minimal Erythemal Dose, MED, which varies by skin phototype).
• Thermal stress: heat-index and wet-bulb globe temperature (WBGT) thresholds from NIOSH/OSHA occupational heat-stress guidance define safe continuous-exposure limits that scale down with physical activity level.
The WHO's 2021 tightening of the annual PM2.5 guideline from 10 to 5 µg/m³ — a 50% reduction — reflects an expanding evidence base linking even low-level chronic particulate exposure to cardiovascular and respiratory mortality, and directly raises the bar that any composite exposome score must be benchmarked against.
A typical composite exposome risk score follows a dose-weighted, reference-normalized additive structure:
Riskᵢ = (measured doseᵢ / reference limitᵢ) × weightᵢ
Composite Score = 100 × Σᵢ Riskᵢ / Σᵢ weightᵢ (rescaled to a 0–100 index)
Weights (wᵢ) can be set uniformly, or informed by relative hazard evidence — e.g. weighting PM2.5 more heavily given its strong cardiovascular mortality evidence base (Global Burden of Disease attributes several million annual deaths to ambient and household PM2.5). A ratio above 1.0 for any single axis (e.g. noise dose 1.3× the OSHA action-level-equivalent) indicates that axis alone has already crossed its reference threshold for the day, independent of the composite.
This structure mirrors established occupational multi-hazard indices and lets the wearable flag not just an overall number, but which specific axis is driving elevated risk on a given day — actionable in a way a single undifferentiated score is not.
A single day's composite score has limited predictive value in isolation; the exposome framework — as originally articulated by Wild (2005) and expanded by the NIH and EU HEALS consortium — emphasizes that health risk accumulates from the general external exposome (broad environmental and social factors), the specific external exposome (individually measurable exposures like this wearable captures), and the internal exposome (biological response, captured partly by Stage 3's biomarkers).
A wearable-derived composite score is therefore best interpreted as a rolling, multi-day trend rather than a single-day verdict — chronic elevation across weeks is far more consequential for long-term disease risk than a single high-noise commute or one high-UV afternoon outdoors, even though both register instantaneously on the dashboard.
The exposome wearable's value is realized only when data becomes action. Closing the loop — route rerouting around high-pollution corridors, automated air-purifier triggers, ventilation nudges, and behavior prompts — converts a passive monitoring device into an active exposure-reduction tool, with the composite risk score serving as the longitudinal outcome metric that shows whether interventions actually work.
Once the composite score identifies which axis is driving elevated exposure, the platform can trigger specific, targeted interventions:
• Route rerouting: using crowd-sourced or municipal air-quality sensor networks alongside the wearer's own historical PM2.5 traces, the app suggests walking/cycling/driving routes that avoid high-traffic corridors during peak commute windows — published personal-exposure studies have found low-pollution routing can reduce inhaled PM dose by roughly 20–30% for the same origin-destination trip, sometimes with minimal added travel time.
• Automated air purifier triggers: when indoor PM2.5 crosses a threshold (e.g. from cooking or wildfire infiltration), a smart-home-linked purifier is triggered automatically; a well-sized HEPA unit run continuously can reduce indoor PM2.5 by roughly 40–60%.
• Ventilation alerts: CO2/VOC buildup in a bedroom or office prompts a nudge to open a window or run mechanical ventilation, addressing both particulate dilution and the well-documented cognitive effects of elevated indoor CO2.
• Behavioral nudges: noise dose approaching the OSHA/NIOSH action level prompts a suggestion to lower media volume or don hearing protection; high UV Index prompts sunscreen/shade reminders timed to accumulated SED rather than clock time alone.
A single day's intervention is a minor perturbation; the meaningful signal is whether the composite exposome risk score trends downward over weeks as interventions compound and become habitual. Longitudinal personal-exposure tracking — the same principle underlying long-running cohort studies but now individualized and continuous — lets a wearer (or a research cohort) see:
• Baseline variability before any intervention (natural day-to-day and weekday/weekend swings) • A step change immediately after a specific intervention (e.g. a new commute route) • Gradual drift as habits solidify or lapse
This before/after, trend-based framing is exactly how environmental health researchers evaluate exposure-reduction interventions in the field — for example, panel studies evaluating clean-cookstove distribution or urban low-emission zones track biomarker and personal-monitor trends over weeks to months, not single-day snapshots.
In this simulated 14-day panel, layering route rerouting, purifier automation and behavioral nudges progressively lowers PM2.5 dose by ~34%, VOC index by ~21% and noise dose by ~18%, moving the composite exposome risk score from 58/100 down to 41/100 — illustrating how multi-axis, sustained intervention outperforms any single fix.
Aggregated across many wearers, individual exposome streams become a research asset far beyond any one person's dashboard. This is the ambition behind large-scale programs like the EU's HEALS project and the broader push toward an NIH-style Human Exposome Project: fleets of low-cost wearables generating dense, geolocated, multi-channel exposure data that fixed-site regulatory monitoring networks — by design sparse in space and blind to individual mobility — cannot provide.
Combined with biomonitoring (Stage 3) and health-outcome records, population-scale wearable exposome data offers the possibility of finally closing Christopher Wild's original 2005 challenge: measuring the environmental half of disease risk with the same rigor, resolution and scale that genomics has achieved for the inherited half.