Extreme Heat and Human Health: The Latest Evidence

Across 732 locations in 43 countries, about 37% of warm-season heat-related deaths between 1991 and 2018 could be attributed to human-caused warming. That figure comes from a multi-country attribution analysis published in Nature Climate Change in 2021, and it has held up well. It changed how I talk about heat. Heat-related mortality is no longer a projection to be modeled for 2050. A measurable share of it is already here, already counted, and already ours.

Mortality is the easy part, though, at least methodologically. The harder questions are which organ systems fail, through which pathways, in whom, and how strongly the evidence supports each claim. Those are the questions I want to work through here. I've sorted the evidence by how much weight it can bear, because the heat-health literature mixes very robust findings with plausible but thin ones, and the papers don't always make the difference obvious.

The physiological starting point

The core mechanism isn't controversial. To shed heat, the body redistributes blood to the skin and sweats. Cutaneous vasodilation demands higher cardiac output. Sweating depletes plasma volume. Under sustained heat, that combination raises heart rate and myocardial oxygen demand, concentrates the blood, increases viscosity and pushes toward a prothrombotic state, while reduced renal perfusion and volume depletion stress the kidneys.

A healthy young adult can usually compensate. Someone with heart failure, chronic kidney disease, impaired sweating, or a medication that interferes with any of these responses often can't. Most of what follows is that principle applied organ by organ.

Physiological limits have also been revised. The widely cited theoretical survivability threshold of a 35°C wet-bulb temperature was always an upper bound for an idealized person at rest. Controlled chamber studies in young, healthy volunteers have since suggested that the critical environmental limit in hot, humid conditions sits closer to 31°C wet-bulb, and lower still in hot, dry conditions. For older adults it is almost certainly lower again. I'd treat 35°C as a figure for climate physics, not for public health planning.

What is established: meta-analyses and large multi-country studies

All-cause and cardiovascular mortality

The strongest evidence comes from large time-series and case-crossover studies, particularly the Multi-Country Multi-City (MCC) Collaborative Research Network and comparable consortia. They consistently show a nonlinear relationship between temperature and mortality. Risk is lowest at a location-specific minimum mortality temperature and rises steeply into extreme heat, usually within a lag of zero to three days.

Cardiovascular deaths account for a large share of that excess. Multi-country analyses find higher cardiovascular mortality on extreme heat days, with consistent signals for ischemic heart disease, stroke and heart failure. Heart failure stands out to me mechanistically. It is the condition in which the extra cardiac output demanded by thermoregulation is least available.

One nuance belongs in any honest summary. Earlier global analyses found that more temperature-attributable deaths were associated with cold than with heat, and with moderate rather than extreme temperatures. That doesn't make heat a minor problem. Heat effects are acute, concentrated and increasing, and they are more tractable to short-term warning systems. But anyone who writes as if non-optimal temperature means only heat will be called out by reviewers, and rightly so.

You can follow this literature through a PubMed search on extreme heat and cardiovascular mortality.

Pregnancy and preterm birth

A 2020 systematic review and meta-analysis in The BMJ pooled studies across many countries. It found that the odds of preterm birth rose with higher ambient temperature and were elevated during heatwaves. Associations with stillbirth were also reported. The direction of effect is consistent across settings, and I'd now call the preterm birth association established at the population level.

The mechanisms are less settled. Candidate pathways include dehydration-related reductions in uterine blood flow, heat-induced release of oxytocin and prostaglandins, inflammation and placental stress. These are plausible. Few of them have been demonstrated in human studies with the precision the epidemiology would justify.

Kidney outcomes

Hospital admission and emergency department studies consistently show that high temperatures increase presentations for acute kidney injury, kidney stones, dehydration and electrolyte disorders. Of all organ-specific signals, the renal one is among the most reproducible in morbidity data, and I think it is still underappreciated outside nephrology. Heat-related kidney morbidity often appears in admission records rather than death certificates, so mortality-focused studies miss much of it. A PubMed search on ambient heat and acute kidney injury is a good entry point.

Strong but more contested: observational and mechanistic signals

Mental health

The mental health evidence has grown quickly and points in a consistent direction. Higher temperatures are associated with more emergency presentations and admissions for psychiatric conditions, higher suicide rates, and more interpersonal violence. Meta-analytic work has reported increased mental health-related mortality and morbidity per degree of warming. Large analyses of suicide data from the United States and Mexico have found that suicide rates rise with monthly temperature.

I place this a tier below cardiovascular mortality for a few reasons. Outcome definitions vary widely across studies. Exposure-response shapes are less consistent. Confounding by season, daylight, holidays and service availability is harder to rule out. The mechanisms are also more speculative. Sleep disruption is the explanation I find most persuasive, because the link between night-time heat and sleep loss is well documented and sleep loss affects mood and impulse control. Proposed changes in serotonergic signaling are biologically reasonable but supported mostly indirectly.

One clinical point is solid. People who take psychotropic medication face pharmacological risks in heat. Antipsychotics and anticholinergic drugs can impair sweating and thermoregulation. Lithium has a narrow therapeutic window, so dehydration can tip patients into toxicity. That isn't an emerging finding. It is basic pharmacology that heat-health plans still too often leave out.

Chronic kidney disease of uncertain etiology

This is the most interesting open debate in the field, in my view. Epidemics of chronic kidney disease not explained by diabetes or hypertension have been described among agricultural workers, especially sugarcane cutters in Central America, and in parts of South Asia and elsewhere. The heat-stress hypothesis proposes that repeated episodes of heat exposure, dehydration and subclinical kidney injury accumulate into chronic disease. Hyperuricemia and fructose-vasopressin pathways have been suggested as contributors.

The hypothesis is plausible and has evidence behind it. Workplace interventions emphasizing water, rest and shade have been associated with less kidney function decline. But competing and complementary explanations remain in play, including agrochemical exposures, heavy metals, infections and nephrotoxic medications. I don't think this is a single-cause disease. The most productive work treats occupational heat as a necessary but probably not sufficient component. See the literature on Mesoamerican nephropathy and heat stress.

Emerging and early evidence

Several lines of work are worth tracking but are not yet strong enough for confident claims:

• Gestational diabetes and hypertensive disorders of pregnancy. Some cohort studies report associations with heat exposure. Findings are inconsistent, and exposure windows differ in ways that make pooling difficult.

• Cognition and learning. Studies link heat to poorer test performance and reduced cognitive function, including in students without air conditioning. These are interesting, but heterogeneous designs leave the effect sizes uncertain.

• Animal models of recurrent heat stress. Rodent work has shown kidney injury after repeated heat exposure with dehydration, and has helped generate hypotheses about the fructose-vasopressin pathway. As always, translating thermoregulatory physiology from rodents to humans calls for caution.

• Heat and medication interactions beyond the obvious. Diuretics, beta-blockers and anticholinergics are well recognized. Newer drug classes that affect volume status or sweating deserve systematic pharmacoepidemiologic study, and much of that study hasn't been done yet.

Who is most at risk

Vulnerability is better characterized than most mechanisms. The groups with the most consistent evidence are:

• older adults, especially those living alone or with limited mobility

• people with cardiovascular, respiratory or kidney disease, or diabetes

• people with serious mental illness or dementia, and those on psychotropic medications

• pregnant people and young infants

• outdoor and manual workers, including agricultural and construction workers

• people experiencing homelessness

• people in low-income urban neighborhoods with limited tree cover and more heat-retaining surfaces

The 1995 Chicago heat wave remains the standard reference on social vulnerability. Social isolation, neighborhood conditions and access to cool spaces shaped who died at least as much as individual clinical risk did.

My concern is that vulnerability research leans heavily on area-level indicators. Ecological variables are useful for planning, but they can't tell a clinician which patient needs a phone call before a heat event. I'd like to see more linked individual-level data on medications, comorbidities, housing and cooling access.

Limitations that should shape how we read this literature

Here is a short version of the methodological issues I'd flag for any student entering this area.

Interventions: where the evidence is thinner than the confidence

Air conditioning is consistently associated with lower heat mortality in observational studies. It is energy-intensive and inequitably distributed, and it increases emissions when grids depend on fossil fuels.

Electric fans are more contested. At very high temperatures and low humidity, especially in older adults with reduced sweating, fans may add heat rather than remove it. Laboratory studies have refined guidance in this area, and public messaging hasn't always caught up.

Heat-health action plans, promoted by the World Health Organization and national agencies, are sensible. Heat mortality in some countries fell after plans were introduced. Isolating the effect of a plan from secular trends, though, is hard with the designs most evaluations use. I'd like to see far more rigorous evaluation, including stepped implementation designs where they're ethical. Trials of cooling interventions, hydration protocols and occupational measures are slowly appearing. Searching ClinicalTrials.gov for heat-related interventions shows how small that pipeline still is compared with the observational literature.

Pharmacological prevention of heat illness is essentially absent as a development area. The practical levers are medication review before summer, sick-day rules for drugs affecting volume status, and clear guidance for patients on lithium and antipsychotics. These are unglamorous measures. I suspect they would prevent more harm than most of what gets funded.

What I'd watch next

These are the developments I'm paying attention to:

1. Indoor and personal exposure measurement. Wearables and low-cost indoor sensors could finally fix the exposure misclassification problem. The first well-designed cohorts using them will reset some effect estimates.

2. Morbidity burden estimates. Mortality dominates the headlines. Admissions, kidney injury, adverse pregnancy outcomes and lost work capacity probably represent a larger total burden, and better accounting would change policy priorities.

3. Causal clarity on CKD of uncertain etiology. Longitudinal cohorts that measure heat, hydration, agrochemical exposure and kidney biomarkers together are the only realistic way to resolve this debate.

4. Evidence from low- and middle-income countries. Expect better data from South Asia, sub-Saharan Africa and Latin America as surveillance improves. Expect the findings to challenge exposure-response curves borrowed from Europe and North America.

5. Rigorous intervention trials. Every trial of cooling, workplace protection or heat warnings matters, because this field has far more associations than tested solutions.

The literature now spans cardiology, nephrology, obstetrics, psychiatry, occupational health and climate science, and it rarely cross-references well. When I map a field like this, I keep the searching, reading and synthesis in one place so the threads don't get lost. That's what an AI-native research workspace like PACR's research search is built for: you can pull the cardiovascular, renal and perinatal evidence into the same project, summarize it, and draft from it without starting over each time. If you're beginning a review or a grant section on heat and health, you can set up a PACR workspace and try it on this topic.

My own suggestion is simpler. Pick one organ system outside your specialty and read its heat literature this month. The most useful ideas in this field keep coming from people who did exactly that.