Airborne Infection Risk Estimator

CO₂-informed SARS-CoV-2 transmission model
Mathematical structure informed by the PMC technical appendix

This is an independent implementation with independent calibration and uncertainty bands. Not affiliated with or endorsed by PMC.

AIRE v3.5 (Last updated February 28, 2026)

Model Assumptions v1.0

Educational tool — not medical advice
1 Community Prevalence

Fraction of population currently infectious. Derived from: (daily cases × infectious window) ÷ population.

≈ 1 in 100 people actively infectious

Optional client-side fetch of a public JSON file hosted at pmc19.com. No user inputs are transmitted. If the fetch fails or you are offline, your manual prevalence is used. The JSON is maintained by PMC19; this tool only reads it. Cached locally for 6 hours.

2 Exposure Setting
1000

Use a stable reading — typically 1–3 minutes after the number stops drifting. Avoid readings taken right after opening a door or window.

15
60

CO₂ reflects room-average air; close-range exposure can be substantially higher.

3 Modeled Risk Estimates
Values shown as Low / Mid / High (uncertainty band) These ranges reflect uncertainty in prevalence, emission intensity, and dose-response calibration. They are structured sensitivity bands, not confidence intervals.
P(infectious person present)
P(your infection this event)
P(≥1 infection per year, if repeated weekly)
Expected infections per year (modeled, if repeated weekly)
P(≥2 infections per year)
P(≥3 infections per year)
Rebreathed air fraction
Event risk level (mid estimate)
Very LowLowModerateHighVery High
CO₂ reflects far-field (room-average) risk. Close-range exposure can be substantially higher, especially with loud speech, crowding, or prolonged face-to-face interaction.
⚠ This estimate does not include cumulative harm from repeated infections. Lower risk is not zero risk. Even clinically "mild" infections can cause lasting biological effects.
What drives the range?
High
Community prevalence. Dominant source of uncertainty. Rarely known precisely — fluctuates with surveillance quality and wave timing.
Mod
Activity / emission intensity. Talking, breathing rate, and individual variation ("super-emitters") affect aerosol dose substantially.
Mod
Close-range (near-field) contact. CO₂ reflects room-average air. Face-to-face exposure is not captured and can be much higher.
Low
Harm assumptions (life-minutes per infection). Population-level estimates vary by age, health status, and variant. Affects cigarette/life-minute outputs only.

Real-world transmission is overdispersed: a small number of individuals can account for a large share of spread. This model estimates average far-field risk under structured uncertainty. It does not simulate super-emitter dynamics.

3b Risk Budget

Expected infections/year (mid estimate) compared to your chosen budget. Based on 52 weekly repetitions at current settings.

4 Comparative Risk (Optional)

Communication metaphor, not biological equivalence. Life-minutes shown first; packs are optional.

i Model Transparency & Limitations

Infectious Prevalence

p = (estimated_daily_infections × infectious_window) / population

Default infectious window: 7 days. The LOW–HIGH uncertainty bands apply prevalence multipliers (see Uncertainty Bands below) to reflect surveillance and timing uncertainty.

Uncertainty Bands (Model Assumptions v1.0)

Loading…

These are structured sensitivity bands — not confidence intervals. They combine prevalence uncertainty (dominant), dose-response calibration uncertainty, and emission/activity uncertainty into three scenarios. The result is a transparent range showing how much outputs could shift under plausible alternative assumptions.

Uncertainty Structure

The Low / Mid / High bands shown throughout this tool reflect structured parameter variation — not statistical confidence intervals. Each output is computed three times under different combinations of the multipliers below.

Loading…

These multipliers are applied simultaneously: the Low scenario uses all low-end values, the High scenario uses all high-end values. This is intentionally conservative at both tails, approximating the plausible range of outputs under joint parameter uncertainty.

These ranges are heuristic and versioned under Model Assumptions v1.0. They are not derived from formal uncertainty quantification and may be revised in future versions.

Rebreathed Air Fraction (CO₂ Mode)

f = (C_indoor − C_outdoor) / (C_exhaled − C_outdoor)
where C_exhaled ≈ 38,000 ppm

Steady-state CO₂ mass balance. Outdoor CO₂ defaults to 420 ppm. Exhaled CO₂ of ~38,000 ppm (3.8%) is a standard physiological value.

Rebreathed Fraction (Room Mode)

Uses conservative pre-estimated fractions by room type, optionally adjusted by a ventilation quality multiplier: Worse than typical = ×1.5, Typical = ×1.0, Better than typical = ×0.7. Applied to the base rebreathed fraction only.

Dose Proxy

D = f × time_hours × activity_mult × crowd_mult × air_cleaning_factor

Activity: Quiet=1.0, Normal=2.0, Loud=5.0. Crowd: Dispersed=0.8, Mixed=1.0, Crowded=1.5. Air cleaning: None=1.0, Typical HEPA=0.7, Strong HEPA=0.5.

Dose-Response Model

P(infection | infectious present) = 1 − exp(−α × D)

α is user-selectable: Conservative=1.5, Moderate=2.8 (default), Higher transmissibility=4.5. Reflects uncertainty in infectious dose-response and variant dynamics. Higher values produce higher infection probability for the same exposure.

Event Probability

P(infectious present) = 1 − (1 − p)^N
P(infection) = P(infectious present) × P(infection | infectious present)

Annual Projections (Four Metrics)

E(infections/year) = 52 × P_event
P(≥1 infection/year) = 1 − (1 − P_event)^52
P(≥2/year) = 1 − [(1−P)^52 + 52·P·(1−P)^51]
P(≥3/year) = 1 − Σ_{k=0}^{2} C(52,k)·P^k·(1−P)^{52−k}

Expected infections is the average count over 52 weekly repetitions. P(≥1) is the probability of at least one infection. P(≥2) and P(≥3) support "risk budgeting" — for example, deciding whether an activity keeps you below 2 infections per year. All assume identical weekly exposure and constant prevalence. P(≥1) saturates toward 100% faster than the expected count grows; P(≥2) and P(≥3) lag behind P(≥1) and better differentiate moderate-risk scenarios.

Cigarette Pack Equivalence (Modeled Range)

life_min_lost = P_event × life_min_per_infection
packs = life_min_lost / minutes_per_pack

Default: 400 minutes per pack (20 cigarettes × 20 min each). Life-minutes-per-infection are modeled by age band and baseline health risk category — rough population-level placeholders, not predictions for individuals. Shown as Low/Mid/High ranges reflecting the same scenario bands applied to P_event. This is a communication aid; it does not imply similar mechanisms of harm between infection and smoking.

Limitations

  • CO₂ captures far-field (well-mixed) risk. Near-field exposure may substantially exceed this estimate.
  • Prevalence is a major uncertainty — rarely known precisely.
  • Assumes well-mixed air; real rooms have spatial gradients.
  • Viral load variation, mask use, prior infections/vaccination, and individual susceptibility are not explicitly modeled.
  • Cigarette equivalence is a communication heuristic, not a precise health measure, and does not imply similar biological mechanisms.
  • Cumulative or compounding harm from repeated infections is not modeled — each event is treated independently. Long COVID is not captured in granular detail.
  • Baseline health risk categories adjust severity-weighting only. "Lower risk" does not mean "safe" — even mild infections can cause lasting biological harm.
  • Not a substitute for professional medical or public health guidance.
  • Transmission is overdispersed (a few cases drive much spread); this tool estimates average far-field risk with structured uncertainty and does not simulate super-emitter dynamics.

Calibration Status

Current parameter values used in scenario bands:

Loading…

Future versions may calibrate α and emission multipliers to benchmark scenarios (e.g., known super-spreader events, classroom transmission studies). This has not been done yet.

Model Assumptions Change Log

v1.0 (initial version)

  • Structured Low / Mid / High uncertainty bands introduced (prevalence, emission, and α multipliers).
  • Prevalence band: ×0.6 / ×1.0 / ×1.6. Emission band: ×0.8 / ×1.0 / ×1.3. α band: ×0.7 / ×1.0 / ×1.4.
  • Annual risk budgeting implemented (expected infections vs. user-selected budget).
  • Life-minutes designated as primary harm output; cigarette packs secondary.
  • Optional PMC19 JSON prevalence fetch with 6-hour local cache (public data only; no user data transmitted). Maintained by PMC19; this tool only reads the published file.