The 2,700 figure is a statistical model estimate of “excess” heat-related deaths, not a count of verified individual deaths.

Oxford researchers (Prof. Carl Heneghan and Dr. Tom Jefferson from the Centre for Evidence-Based Medicine) have questioned it, noting that official Office for National Statistics (ONS) weekly death data for England and Wales show no dramatic spike during the May/June 2026 heatwave periods compared to recent years.

Researchers from Imperial College London, the Met Office, and LSHTM released a rapid analysis estimating ~2,700 excess deaths linked to heat during two heatwaves:

• ~550 in the May event (peaking ~35.1°C).
• ~2,200 in the June event (peaking ~37.7°C).

They attributed ~42% (~1,140) to human-caused climate change making temperatures 3–4°C hotter than they would otherwise have been.

These models are standard in environmental epidemiology (used by WHO, Lancet, etc.) but remain estimates.

Direct surveillance (autopsies, certified causes) undercounts heat contributions, while models provide broader insight. For the best picture, compare modeled excess against raw mortality data where possible.

These are not “heatstroke” deaths recorded on certificates.

The method uses established epidemiological models: historical temperature-mortality relationships (e.g., higher deaths among elderly/vulnerable from cardiovascular/respiratory strain, dehydration) combined with observed temperatures to calculate “excess” above expected baselines, adjusted for trends. Similar approaches have been used for past UK and European events.

Heneghan and Jefferson argue:

• ONS weekly registered deaths (by registration date) show normal variation—no clear excess spike during the heatwave weeks versus the prior decade.
• Models are useful for exploring risks but should be validated against real surveillance data. When they diverge, prioritize data.
• They are not denying heat risks to vulnerable people (e.g., elderly in poorly adapted homes) but emphasize that many such deaths involve care failures, and attribution requires scrutiny.

Caveats on ONS data:

Deaths are often registered days/weeks later (weekends, holidays, coroner cases), so weekly figures can lag or smooth short events like heatwaves. This is a known limitation acknowledged by the Oxford team.

Context on Heat vs. Cold Deaths

• Heat does increase mortality risk, especially in urban areas with high humidity/nighttime temperatures limiting recovery. Vulnerable groups (elderly, those with pre-existing conditions) are most affected.
• Cold still causes far more excess deaths annually in the UK (typically tens of thousands) than heat.
• These rapid attribution studies aim to quantify risks quickly for policy (e.g., cooling, alerts, building adaptations). Critics argue they can overstate for headlines if not transparently validated.

Excess mortality models

Excess mortality models estimate the number of deaths above what would normally be expected during a specific period (e.g., a heatwave), attributing the difference to a factor like extreme heat. These are statistical/epidemiological tools, not direct counts of certified “heat deaths” on death certificates.

These models are useful statistical tools but have notable limitations, assumptions, and potential for over- or misinterpretation. Oxford’s Carl Heneghan and Tom Jefferson (and similar evidence-based skeptics) emphasize prioritizing raw data over models when they diverge.

Models vs. Observable Reality

• Many estimates (like the recent ~2,700) rely on counterfactual modeling (what would have happened without heat/climate change) rather than counting verified deaths. If official ONS data shows no clear spike in weekly deaths during the heat periods, the large, modeled number raises questions.
• Heneghan/Jefferson argue: Surveillance data should ground models. Models explore possibilities; they don’t prove what happened. When they conflict, investigate further rather than headline the model.

“Harvesting” or Mortality Displacement

• Heat (like cold or flu) often advances deaths in frail, elderly, or comorbid people by days/weeks. The excess appears during the event, but total deaths over months may show little net increase (some “borrowed” from future). Models sometimes capture short-term spikes without fully adjusting for this.
• Result: Reported “thousands of heat deaths” can sound alarming but partly reflect timing shifts in already vulnerable populations.

Baseline Sensitivity

• Excess depends heavily on the chosen baseline (e.g., previous 5–10 years, smoothed trends, exclusions like COVID). Different baselines or trend assumptions produce varying results. Long-term declines in mortality or aging populations complicate this.
• Seasonal confounders (allergies, infections, holidays, behavioral changes) are hard to fully control.

Attribution Challenges (Especially Climate Change Portion)

• Separating “heat” from “climate change-added heat” involves climate models + epidemiological models — compounding uncertainties.
• Rapid attribution studies (common now) are useful for speed but less rigorous than full peer-reviewed analyses with more validation time.
• Other factors during heatwaves (e.g., poor urban planning, lack of AC/cooling, dehydration from inadequate care, air pollution, concurrent events) often drive risk more directly than the temperature anomaly itself.

Methodological Issues in Common Approaches (e.g., DLNMs)

• Distributed Lag Non-linear Models are flexible but can overfit or be sensitive to lag choices, temperature metrics (dry bulb vs. humidex), and thresholds.
• They assume stable historical relationships that may change with adaptation (better warnings, AC prevalence, behavioral shifts).
• Some critiques note potential overestimation in projections if adaptation is underestimated.

Broader Context and Incentives

• Heat deaths get more media/policy attention than cold (which still dominates excess winter mortality in the UK/Europe), even if absolute numbers are smaller.
• Funding, media, and policy environments can favor alarming estimates. Heneghan et al. have criticized similar dynamics in COVID excess death reporting.
• Direct certified heat deaths remain low; models fill the gap but aren’t a substitute for transparent, timely raw data by date of death.

Counterpoints (for Balance)

• Raw data does undercount heat contributions (most deaths coded as heart attack/stroke, not “heat”).
• Models align with observed physiology: heat stresses cardiovascular/respiratory systems, especially in cities.
• Trends show rising heat vulnerability with warming + aging populations.

Excess mortality models highlight real risks and can guide adaptation (better homes, alerts, urban greening).

However, they are not precise counts of lives lost “due to heatwaves/climate change.”