Predicting excess-mortality risk from a heatwave using temperature/humidity dynamics and population vulnerability stratification
Heat is the deadliest weather hazard in most countries, yet it is also the most unevenly distributed: two residents experiencing the identical outdoor temperature can face wildly different mortality risk depending on age, housing, social connectedness, and access to cooling. Before any heatwave forecast can be turned into a mortality prediction, a city must first be mapped — its building stock, its population distribution, and the vulnerability characteristics layered across every neighborhood.
Heat vulnerability is not randomly distributed — it clusters along predictable demographic and physiological lines. Age is the strongest single predictor: adults over 65 have measurably reduced sweat gland output, blunted thirst perception, and slower cardiovascular adaptation to heat load, meaning their core temperature rises faster and takes longer to recover once elevated.
Chronic disease compounds age-related risk. Cardiovascular disease patients cannot increase cardiac output enough to redirect blood to the skin for cooling; diabetics often have impaired sweat responses from autonomic neuropathy; patients on diuretics, beta-blockers, antipsychotics, or anticholinergic medications have measurably impaired thermoregulation as a drug side effect.
Social and structural factors matter as much as biology. Living alone removes the informal safety net of someone noticing early symptoms. Top-floor apartments in older buildings without insulation or cross-ventilation trap heat overnight, denying the body the nighttime recovery window it needs. Lack of air-conditioning access — whether from cost, power outages, or housing quality — removes the single most effective individual protective factor identified in heat-mortality studies.
Cities are systematically hotter than their surroundings. Dark asphalt and roofing absorb and re-radiate solar energy; the built environment reduces evaporative cooling from vegetation and soil moisture; waste heat from vehicles, air conditioners, and industry adds a direct thermal load; and dense building geometries trap longwave radiation at night, preventing the cooling that rural areas experience after sunset.
The result is the urban heat island (UHI): city cores can run 1–7°C hotter than surrounding rural land, with the gap often largest at night — precisely when the body needs relief to recover from a hot day. UHI intensity is not uniform within a city either; it concentrates in dense, low-vegetation, low-income neighborhoods that frequently overlap with populations already carrying elevated physiological risk, compounding exposure and vulnerability in the same census tracts.
Historical redlining maps from the 1930s U.S. correlate strongly with present-day urban heat island severity — formerly redlined neighborhoods can run 5°C hotter than formerly greenlined neighborhoods in the same city, a legacy of disinvestment in tree canopy and green space that now compounds directly into heat mortality risk.
Air temperature alone is a poor predictor of physiological heat stress. Humidity determines how effectively the body can lose heat through sweat evaporation — the dominant cooling mechanism once air temperature approaches skin temperature. The heat index (or "apparent temperature") folds temperature and humidity into a single number that tracks much more closely with actual heat-illness risk than dry-bulb temperature does on its own.
The U.S. National Weather Service heat index uses the Rothfusz regression, an empirical multiple-regression fit to a human heat-balance model developed by Steadman (1979):
HI = −42.379 + 2.049·T + 10.143·R − 0.225·T·R − 0.00683·T² − 0.0548·R² + 0.00123·T²·R + 0.000853·T·R² − 0.00000199·T²·R²
(T = air temperature in °F, R = relative humidity in %; the regression applies above roughly 27°C / 80°F — below that, heat index tracks close to air temperature.)
The NWS defines four danger categories from the resulting value: Caution (27–32°C / 80–90°F — fatigue possible with prolonged exposure), Extreme Caution (32–39°C / 90–103°F — heat cramps and exhaustion possible), Danger (39–51°C / 103–125°F — heat exhaustion likely, heat stroke possible), and Extreme Danger (≥51°C / 125°F — heat stroke highly likely). These thresholds are the backbone of most municipal heat-health warning systems.
Heat index is a useful public communication tool, but occupational and physiological researchers often use Wet-Bulb Globe Temperature (WBGT), which additionally accounts for wind speed and solar radiation, making it more accurate for outdoor labor and athletic settings.
A landmark 2010 analysis (Sherwood & Huber) established a theoretical wet-bulb temperature ceiling of 35°C: above this threshold, even a healthy, resting, unclothed human in the shade with unlimited water cannot dissipate metabolic heat fast enough, and core temperature rises uncontrollably regardless of hydration. In practice, dangerous physiological strain begins well below this ceiling — sustained wet-bulb temperatures above 30–32°C are associated with sharp increases in heat illness even among fit adults doing moderate activity.
Humidity is the multiplier that turns a merely hot day into a lethal one: at low humidity, sweat evaporates efficiently and can dissipate substantial metabolic and environmental heat load; above roughly 70% relative humidity, evaporative cooling capacity drops by a majority, which is why humid heatwaves (as in the U.S. Midwest and South, or South Asia) produce disproportionate mortality relative to their peak dry-bulb temperature alone.
The 1995 Chicago heatwave reached a heat index of 48°C (119°F) on its worst day — combined with a strong urban heat island and minimal overnight cooling, it produced an estimated 739 excess deaths in five days, overwhelming the city morgue capacity.
Identical heat exposure does not translate into identical mortality risk. Epidemiological studies of past heatwaves consistently find relative-risk multipliers of 2–6× between the most and least vulnerable subgroups within the same city on the same day. Stratifying the population by risk tier — rather than treating exposure as uniform — is what turns a weather forecast into an actionable mortality prediction.
Heat-mortality models typically express risk as a baseline hazard (driven by heat index) multiplied by a set of individual risk factors that compound roughly multiplicatively:
Risk(individual) = f(HeatIndex) × RR(age) × RR(comorbidity) × RR(isolation) × RR(no-AC)
Each relative-risk term is derived from case-control and time-series studies of past heatwave mortality. Age above 65 alone confers roughly 2–3× baseline risk; cardiovascular or respiratory disease adds a further ~2–2.5×; social isolation (living alone, infrequent visitors) contributes an independent ~2×; and lack of air-conditioning access is consistently the single largest modifiable factor, associated with roughly 80% risk reduction when present.
Stacked together, a socially isolated, cardiovascularly compromised individual over 65 without air conditioning can carry a relative risk 5–8× that of a young, healthy, air-conditioned adult exposed to the exact same outdoor heat index — which is why aggregate temperature forecasts alone are insufficient for targeting emergency response; risk-stratified population maps are needed to direct resources to the households that need them most.
The 1995 Chicago heatwave produced one of the most influential findings in heat-mortality epidemiology: sociologist Eric Klinenberg's post-mortem analysis found that living alone, being homebound, and lacking social contacts predicted death independently of age or medical condition. Many victims were found days after death, alone in top-floor apartments with windows sealed against perceived crime risk — a protective behavior that became lethal in the heat.
This finding reshaped heat-emergency response worldwide: modern heat action plans in Paris, Chicago, and other major cities now maintain "vulnerable persons registries" — opt-in lists of isolated elderly residents whom outreach workers or volunteers proactively check on by phone or in-person visit during heat alerts, specifically to counteract the isolation risk factor that pure temperature-based warnings cannot address.
Klinenberg's analysis found that isolated elderly Chicago residents in high-crime, disinvested neighborhoods died at far higher rates than similarly isolated elderly residents in neighborhoods with strong street-level social ties — demonstrating that community cohesion itself functions as a measurable protective factor against heat mortality, independent of housing quality or AC access.
Heat illness is a physiological continuum, not a binary state. As core body temperature climbs above the normal 37°C set point, the body moves through progressively more dangerous compensatory failures — from manageable heat exhaustion to life-threatening heat stroke — and the elderly, medicated, or cardiovascularly compromised traverse this cascade faster and with less warning than healthy adults.
Heat exhaustion develops as the cardiovascular system struggles to simultaneously perfuse the skin (for cooling) and vital organs (for function). Symptoms include heavy sweating, weakness, dizziness, nausea, and rapid pulse — the body is straining but its thermoregulatory feedback loops are still intact, and core temperature typically remains below 40°C. With rest, fluids, and cooling, most people recover within hours without lasting harm.
Heat stroke is a fundamentally different, life-threatening event: thermoregulation fails outright. Classically defined by core temperature exceeding 40°C together with central nervous system dysfunction — confusion, slurred speech, seizures, or loss of consciousness — heat stroke frequently presents with hot, dry skin as sweating mechanisms fail. Above roughly 41–42°C, cellular proteins begin to denature, triggering a cascade of organ damage: the gut barrier breaks down and leaks endotoxin into circulation, the liver and kidneys can fail within hours, and disseminated intravascular coagulation can develop. Without rapid, aggressive cooling — ice-water immersion is the gold-standard treatment — mortality approaches 80%; with cooling initiated within 30 minutes, mortality drops to roughly 10–20%.
The critical clinical variable is not peak temperature but time-at-temperature: the "cooling rate" achieved in the first hour after heat stroke onset is the single strongest predictor of survival, which is why heat action plans emphasize rapid identification over waiting for ambulance transport.
Cooling the body during heat stress depends on redirecting a large fraction of cardiac output to the skin — in a young healthy adult, skin blood flow can increase roughly eightfold during heat stress, from ~5% to as much as 30–40% of cardiac output. This redirection requires the heart to substantially increase output, since blood is being sent to the skin in addition to (not instead of) vital organs.
Aging reduces maximal cardiac output, arterial compliance, and baroreflex sensitivity, meaning older adults simply cannot achieve the same skin blood flow increase without dropping blood pressure to unsafe levels. Pre-existing cardiovascular disease compounds this further — a heart with reduced ejection fraction or coronary artery disease may be unable to meet the heat-driven demand at all, leading to the well-documented spike in myocardial infarctions and heart failure decompensation seen during heatwaves, independent of any direct heat-stroke diagnosis.
Common medications sharpen this vulnerability: diuretics reduce blood volume available for the compensatory response; beta-blockers blunt the heart-rate increase needed to raise cardiac output; and anticholinergic drugs (common in dementia and Parkinson's treatment) directly suppress sweating. A patient can be on several such medications simultaneously for unrelated chronic conditions, each independently narrowing their physiological safety margin during a heatwave.
The purpose of a heat-mortality risk model is not academic — it is to trigger action before deaths occur. Modern heat-health early warning systems combine meteorological forecasting, population vulnerability mapping, and predicted excess-mortality curves to activate interventions — cooling centers, outreach calls, utility shutoff moratoriums — days ahead of the peak heat, and the evidence shows these interventions measurably bend the mortality curve.
The 2003 European heatwave remains the deadliest extreme-heat event in modern developed-world history: sustained temperatures 10°C above seasonal norms across Western Europe produced an estimated 70,000 excess deaths, with France alone recording roughly 15,000 — disproportionately among elderly residents in un-air-conditioned apartments during the country's traditional August vacation period, when many normal social-support and medical staffing structures were unavailable.
The 2021 Pacific Northwest heat dome demonstrated that extreme heat mortality is not confined to historically hot regions: temperatures reached 47°C (116°F) in Portland and 49.6°C (121°F) in Lytton, British Columbia — a region with historically low air-conditioning penetration (fewer than half of Seattle and Portland households had AC at the time) — producing roughly 1,400 excess deaths across the region in a matter of days, overwhelming emergency medical systems built for a temperate climate.
Both events share a common signature: mortality concentrated heavily among the elderly, the socially isolated, and those without air conditioning — the same risk factors identified in Chicago 1995 — and both catalyzed major policy reform in heat-health early warning infrastructure.
Modern Heat-Health Early Warning Systems (HHWS) integrate weather-service forecasts with locally calibrated heat-mortality thresholds — the temperature or heat-index level at which a given city's historical mortality data shows excess deaths beginning to rise. When a forecast crosses that threshold with 3–5 days of lead time, a graduated response activates: public alerts, extended cooling-center hours, proactive outreach calls to vulnerable-persons registries, utility shutoff moratoriums, and in the most severe cases, door-to-door wellness checks.
Evaluations of mature heat action plans (Paris, Philadelphia, Ahmedabad) find mortality reductions on the order of 30–50% relative to the pre-program era for equivalent heat exposure, driven primarily by cooling-center utilization and outreach-triggered relocation of the highest-risk individuals during peak danger hours. Cooling centers work through a simple mechanism: even a few hours of exposure to air-conditioned space per day meaningfully lowers cumulative core-temperature strain and allows overnight recovery that a sealed, un-air-conditioned apartment cannot provide.
The remaining challenge is uptake: the populations at highest risk — isolated, mobility-limited, distrustful of outreach, or unaware of alerts — are also the hardest to reach, which is why the most effective programs pair passive public messaging with active, name-based outreach to pre-identified vulnerable registries rather than relying on self-selected cooling-center visits alone.
Ahmedabad, India's Heat Action Plan — launched after a 2010 heatwave killed over 1,300 people in the city alone — is credited with an estimated 25–30% reduction in heat-related mortality in subsequent comparable heatwaves, becoming a template adapted by over 30 cities across South Asia.