Expert Q&A: Why can heat death counts differ so much for the same location?

Published: July 20, 2023

Global Heat Health Information Network


When it comes to estimating the number of heat-related deaths that occur during a heat  season, there are often discrepancies across different studies and reports, even for the same location. These differences can cause confusion and make it difficult to assess the difference in impacts between locations and from year to year.


Last week, research was released estimating 61,672 heat-related deaths occurred across Europe in summer 2022. The research by Ballester et al. is important because it helps provide a consolidated and standardised measure of the impact of extreme heat across the entire European region. However, for individual countries, the analysis calculated figures which may differ from nationally recorded statistics. For example, the study identified 4,807 heat-related deaths in France, but Public Health France’s own analysis put the number at 6,969 – a difference of 2,162 lives. 


We spoke with experts from the Global Heat Health Information Network to find out why these differences between estimates arise, what it means for risk communication, and what the numbers tell us about how heat risks are changing. 

Embed from Getty Images



  • There is no perfect and absolute way to measure heat-related deaths. Results depend on the approach used and the availability and quality of data. 
  • While the limitations of each approach should be made clear, none of the varied approaches commonly used are necessarily wrong. They are each valid scientific explorations of heat impact that provide important insights for planners and policymakers.
  • Even where differing estimates occur, heat-related death counts are generally considered by experts to be gross underestimates of the true impact. There is no dissent in the scientific community that heat is a significant and largely unmanaged risk.   
  • This focus on the difference in estimates distracts from the fact that all the numbers are unacceptably high, particularly as we know that there are effective actions that can be taken to prevent them. 
  • The scientific community’s main challenge is not to improve the precision of heat-death calculations, but to identify and assess solutions to reduce mortality and other heat impacts..
  • Various mortality estimates all point to exceptionally high levels of excess mortality resulting from heat. Such consensus constitutes sound evidence that heatwaves exact a heavy toll on society and forms a compelling case for the appropriate allocation of resources needed to prepare for and address the impacts of heat. 
  • Taking into account the magnitude of heat-related mortality on the continent, the results of Balester et al. call for national governments and relevant agencies in the European Union and continental levels to increase the ambition and effectiveness of heat prevention and adaptation plans with urgency.



Mathilde Pascal (MP), Santé Publique France
Mathilde is an epidemiologist, responsible for the climate change and health program, at the environmental and occupational health department of Santé Publique France. She contributed to the creation of the French heat warning system.


Sari Kovats (SK), London School of Hygiene and Tropical Medicine
Sari is an Associate Professor in the Department of Public Health, Environments and Society in the Faculty of Public Health and Policy. She is knowledge mobilisation lead for the NIHR Health Protection Research Unit in Environmental Change and Health, has been researching the effects of weather, climate and climate change on human health for more than 20 years and has published widely on this topic.


Andreas Matzarakis (AM), Research Centre Human Biometeorology, German Meteorological Service
Andreas Matzarakis is leading the Research Centre Human Biometeorology of the German Meteorological Service in Freiburg and responsible for the Heat Health Warning system in Germany and in general for the biosynoptical weather forecast. He is an extraordinary Professor at the University of Freiburg. His interests are in general in Human Biometeorology and Urban Climatology. He is Developer of several models and tools in applied climatology and biometeorology i.e. RayMan Model, SkyHelios Model and CTIS (Climate-Tourism-Information-Scheme).


Glenn McGregor (GM), Durham University, United Kingdom
Glenn McGregor is a Professor of Climatology and Principal of Ustinov College at Durham University, in the United Kingdom. Glenn’s research explores the relationship between atmospheric circulation and surface environmental processes and the extent to which weather patterns, air mass types and modes of atmospheric circulation (e.g. ENSO, NAO) might influence the intra-seasonal to inter-annual variability of health outcomes. He has particular interests in the causes and consequences of and responses to heatwaves.


Emer O’Connell (EO), Greater London Authority
Emer is an NHS-trained Consultant in Public Health with expertise in the interface between the environment and population health, particularly in relation to inequalities. In her previous role she led the Extreme Events and Health Protection team in the UK Health Security Agency, including the design and delivery of the early warning systems for heat and cold, and was also the focal point for adaptation within the organisation. In her current role, she provides public health expertise on planning, environment and resilience for the London Mayor.


Joy Shumake-Guillemot (JSG), World Meteorological Organization
Joy leads the WMO-WHO Joint Office for Climate and Health in Geneva, Switzerland, and is the co-coordinator of the Global Heat Health Information Network. She is an environmental health scientist and public health practitioner who has worked with WHO, WMO, UNICEF and others to develop public health policy and programming for climate adaptation and risk management. 



First off, why is it important to calculate deaths from heat?


JSG: Calculating who dies from exposure to heat  provides vital information for public health planning, risk assessment, policy development, climate impact monitoring, and raising public awareness in order to develop and take actions. Knowing who is impacted most helps prioritise resources, prevent future deaths, and build resilience in the face of rising temperatures and heatwave events.




Could you explain why different analyses come out with such different numbers when trying to calculate how many people died during a particular heat season or event?


GM: The reason why estimates of heat-related deaths differ is because different studies use different approaches to calculate the numbers. The approaches, while methodologically robust, can vary in terms of the data used, the criteria considered, and the statistical models employed. 

Some studies focus solely on direct heat-related deaths (e.g. where the cause of death was registered as heat stroke), while others consider excess all-cause mortality. This means the deaths above what might be expected for the period of analysis, and includes deaths from indirect heat effects related to heat sensitive diseases such as cardiovascular, respiratory or renal disease. 


SK: Heat deaths can be estimated statistically in terms of the impacts on all cause mortality for days above a heat threshold (usually temperature) over a heat season, and also for individual heatwave events. 


It’s worth noting that there’s no single correct answer, as heat related mortality covers a range of outcomes. It can be used to describe estimates for an average year, or for a specific hot year for temperatures above a threshold (this is similar to how air pollution impacts are estimated). Many heat-related deaths also occur on hot days outside a recognised heatwave event. 


The estimates of individual heatwaves are also likely to vary, because impacts across events and locations will differ significantly depending on many factors, including population exposure, vulnerability and behaviour.


Changes in impact patterns can be more interesting for researchers and policy makers than total mortality figures, for example looking at the locations or age groups most affected. 


JSG: The availability and quality of data also play a crucial role in estimating heat-related deaths. Some countries may have robust health surveillance systems and comprehensive reporting mechanisms, resulting in more accurate data. However, in many areas, there may be underreporting or misclassification of heat-related deaths, leading to an underestimation of the true impact. Incomplete data can skew the estimates and contribute to the disparities observed. 


Studies like Ballester et al. are robust in their approach and helpful to provide a standardised measure of what is happening across different countries, which individually, may use different methods to measure this impact.


MP: In France, two approaches are used to assess the impact of heat on mortality. Firstly, excess deaths during heatwaves are estimated by comparing observed mortality with expected mortality. Calculations of excess mortality provide quick and easy, but do not reveal the specific causes of the excess deaths.


Secondly, models are used to estimate the fraction of heat-related deaths during summer. These models consider various factors and have shown that approximately 33,000 deaths in France from 2014 to 2022 were due to heat, with the highest impact in 2022. These findings align with similar conclusions reached by other research teams.


When comparing estimates, it’s important to consider the temporal precision of the data. The risk of mortality from heat is highest within 0 to 3 days of exposure, so it’s best to have daily data. However, if daily data is not available, using weekly data can still provide some useful insights, although it tends to underestimate the true impact.  


For example, the European Observatory on Climate and Health provides national estimates of heat-related mortality for 2014-2020, based on weekly temperature and mortality data. When we compare their results with ours for France, we found differences ranging from 7% to 60%, depending on the year. Differences are larger during summer with more daily variations in temperatures. This is also the main explanation for the difference in numbers estimated by the Ballester et al. study..


Regardless of the variations that result from methodological differences, all these findings highlight the need to adapt and prepare for heat-related risks and take action to mitigate climate change.


EO: Most heat-related deaths are due to an exacerbation of underlying health conditions, so the death is not likely to be recorded as heat-related through the deaths registration process in most countries. Therefore, we can only use indirect methods such as those used in France to produce estimates of the impacts of heat. 


Ideally, a number of different methods are used, as each approach has strengths and weaknesses. For example, estimating the number of excess all-cause deaths during a heatwave provides a reasonably simple measure of the wider impacts. It is an approach used across other risks, including COVID-19, so also lends some consistency. However, where there is a delay in availability of deaths data and a need for more rapid assessment – for example, to support policy decisions and action – it may be helpful to use the evidence from previous episodes to develop an epidemiological model. This model can be applied to the heatwave scenario to provide estimates of the likely impacts. Applying both of these methods to a single heatwave event will produce different estimates of the impacts, although both estimates are ‘valid’. 


Depending on the strengths and weaknesses of the approach, different users may have a preference for one measure over another. For example, estimates of excess all cause deaths are a useful measure for public health organisations, as it is a metric also used for other public health risks and in some countries these data may be available quickly. However, individual measures are rarely considered in isolation and generally other evidence and supporting data will also be used to develop a wider understanding of the impacts, for example, on healthcare service use.


Differences in estimated impacts that use the same method for seemingly very similar heat events also highlight the dynamic nature of heat risk, of which exposure is just one aspect. Other factors, such as behaviour and the role of concurrent risks like air pollution, also play an important part. Using different methods to estimate the impacts helps us to understand some of the drivers behind these differences. 


AM: Heat is a complex issue. It’s not only air temperature that is important, but also air humidity, wind conditions and exposure to solar radiation. Other factors such as indoor conditions – especially during the night – and specific conditions in urban areas because of the urban heat island, which is more pronounced during the night, are relevant. Other factors such as UV-radiation, ozone or particles in the air, wildfires and pollen can also play a role in how heat impacts health.




Who should the public and policy-makers believe when there are conflicting calculations?


EO: There are limitations with every scientific process, so we must get better at communicating those uncertainties as well as their relevance for policy. High quality peer-reviewed publications and government agencies are more open and transparent about the limitations of their estimates, and the best of these sources will also consider how their findings should be used or applied. Where we fail to do this, discrepancies can lead to the validity of the estimates to be questioned. This focus on the difference in estimates distracts from the fact that all the numbers are unacceptably high, particularly as we know that there are effective actions that can be taken to prevent them. 


It’s also important to note that most of the methods used only consider deaths during periods with the highest temperatures (ie the severe/extreme heatwaves) whereas we know heat-related impacts start to occur at much lower temperatures – so, again, the important message is that all of these estimates likely understate the burden of harm. 


GM: Notwithstanding contrasts in estimates from different studies, the public and policy makers should have faith in the scientific process. For the summer 2022 heatwave events across Europe, the various mortality estimates all point to exceptionally high levels of excess mortality. Such consensus constitutes sound evidence that heatwaves exact a heavy toll on society. That evidence forms a compelling case for the appropriate allocation of resources needed to prepare for and address the impacts of heatwaves. 


If the severity and extent of the ‘heatwave problem’ is underestimated, it may lead to insufficient investments in interventions, emergency response planning, and healthcare infrastructure. The same applies to other heat sensitive sectors such as transport, energy and business. Added to all of this, improved ‘heat risk governance’ is required so that the roles and responsibilities of government departments and agencies and public and private organisations in the development of heatwave response policies, actions and communication can be clearly defined.



Are there any ways to improve the accuracy of heat-related mortality  estimates?


GM: Yes, there are several ways to enhance accuracy. We need common protocols for the statistical methods to ensure proper comparability between studies, and to develop methods to quantify heat impacts in populations that lack high quality daily mortality and morbidity data. It is important that clinicians and others are able to recognise where heat is a cause or contributing factor in a death. Integrating multiple data sources and utilizing advanced modeling techniques can also contribute to more accurate estimates.


MP: In recent years, models have largely improved thanks to a global effort of the scientific communities. Heat-related estimates based on state-of-the-art models are robust. Even simple approaches such as a comparison between observed and expected deaths can provide a good order of magnitude of the heat impact.


However, in many places of the world those models or simpler estimates are not available, resulting in a large underestimation of the heat-related mortality world wide. So, today, efforts should be dedicated to building capacity in those areas. 


From a research point of view, the challenge today is also not to improve the precision of heat-death calculations, but to take into account synergies between risks (i.e. what are the joint impacts of heat and forest fires), and above all, to identify solutions to reduce the mortality impacts.



Learn more: Heat and Health