European Union flag

At the European Environment Agency (EEA), we are regularly asked this question, particularly when we publish annual data on the health impacts of air pollution in Europe. Our answer usually explains our working methods and the strength of existing data, but given how often the question is asked, we would like to provide a more detailed answer.

The question itself is valid. As mortality is primarily recorded through death certificates, it is assumed that deaths caused by air pollution are measured by counting how many certificates list it as the cause.

What is in a death certificate

Death certificates (technically known as ‘medical certificates of cause of death’) are not designed or intended to record information like the exposure to air pollution (see Figure 1).

Figure 1. A standard medical certificate for cause of death from the World Health Organisation

Please select a resource that has a preview image available.

Let us examine the structure of the document. Aside from the basic administrative data, the certificate has two main sections: Frame A, which contains the main medical data related to the death; and Frame B, which records other medical data.

Frame A is subdivided into two parts (see Figure 2). Part 1 is used for diseases or conditions that form part of the chain of events leading directly to death. The immediate (direct) cause of death is entered in line 1(a), and on the other lines, the certifying doctor works backward through earlier conditions to determine what started the chain of events (the underlying cause of death).

Figure 2.  An example of ‘Frame A’ from a medical certificate by the World Health Organization

Please select a resource that has a preview image available.

As can be seen, all the information in this section is meant to be purely medical, as coded in the International Classification of Diseases. Therefore, risk factors like tobacco, asbestos or exposure to air pollution would not be written here.

Could air pollution then be listed in ‘Frame B: Other medical data’ (see Figure 1)?

Frame B has five subsections: previous surgery, manner of death, place of occurrence, fetal or infant death, and pregnancy status (if the deceased was a woman). Of these, only ‘manner of death’ might include non-medical causes or risk factors. However, most of the listed options — excluding ‘Could not be determined’, ‘Pending investigation’ or ‘Unknown’ — require direct verification by the certifying doctor, either through an examination of the body or through an external report. In most cases, the certifying doctor cannot list significant non-medical risk factors (like air pollution) that may have contributed to the person’s death. This information is simply not available to them.

Similarly, death certificates do not contain information on other well-known risk factors for mortality, such as diet, alcohol consumption or smoking. When the Office for National Statistics in the United Kingdom (UK) received a public inquiry about whether it had information on deaths caused by smoking, it responded with:

‘[Our] mortality data comes from the information collected at death registration. All the conditions mentioned on the death certificate are coded using the International Classification of Diseases, Tenth Revision (ICD-10). Unfortunately, … smoking habits are not recorded as part of this process.’ 

Pollution data is also not recorded as part of this process and therefore, death certificates usually do not list air pollution as a cause of death.

Is there such a thing as an exceptional death certificate?

We say usually because there has been at least one exceptional certificate. In 2013, a nine-year-old girl called Ella Adoo-Kissi-Debrah died following an asthma attack and became the first person in the UK, and possibly in the world, to have air pollution listed as a cause of death. The girl lived near a very busy road in southeast London, in an area with high pollution levels. The UK government’s coroner ruled that ‘[a]ir pollution was a significant contributory factor to both the induction and exacerbations of her asthma’. He concluded that Ella had died of asthma, made worse by prolonged exposure to excessive air pollution. The medical cause of death was listed as:

  • 1a) Acute respiratory failure
  • 1b) Severe asthma
  • 1c) Air pollution exposure.

This report linking air pollution to an individual death was only possible after a legal journey that took more than 5 years. The coroner also had access to an overwhelming amount of information, and had the mandate, time and resources to examine it. In a subsequent report following the inquest into her death, the coroner stated that ‘[a]ir pollution was not discussed as a possible causative factor (throughout Ella’s illness) even though Ella was seen by [general practitioners] and specialists in several specialist hospitals. It is an issue that needs to be raised with […] organisations responsible for medical education’.

Seeing the lengthy, costly and burdensome process that led to this outcome, it seems unlikely that death certificates will routinely begin to include air pollution (or other environmental risk factors such as noise pollution, water pollution or exposure to chemicals) as a cause of death.

Therefore, in the absence of such information on death certificates, we count on epidemiological research to find out how air pollution affects health, and how it may increase the risk of premature death in various populations.

Though it has long been observed  that an increase in air pollution is followed by rises in deaths, the first coherent studies linking air pollution and mortality date back to the 1950s. Those studies prompted many industrialised countries to pass their first environmental and clean air laws in the decades that followed. The first large and systematic cohort studies of mortality and air pollution were published in the mid 1990s. Over the past 30 years, these studies have been extensively and independently reanalysed, their results replicated in several other cohorts worldwide and their findings confirmed through extended research. In summary, there has been epidemiological evidence linking  air pollution to mortality for over 70 years, and a systematic observation thereof, in cohorts across various continents and millions of people for the past 30 years.

These cohort studies follow defined groups of people over time and collect data on almost every known health risk factors (age, sex, ethnicity, smoking status, lifestyle, etc.). They then usually compare an exposed group of individuals to an unexposed group. In the case of air pollution, since everyone is exposed to it to a certain extent, these studies compare groups exposed to different levels of air pollution, often in different locations. Over time, some people in the groups will inevitably develop health conditions — some of which are already known — thanks to laboratory and other scientific studies — to be linked to air pollution exposure. By statistically controlling for the other known factors that could also cause these health conditions, researchers can estimate how much air pollution contributes to these outcomes. They can also determine how that contribution of pollution changes at different levels of exposure. Using this information, they build exposure-response functions (or curves) — mathematical models that link different levels of air pollution to different levels of risk of developing certain health conditions or of dying prematurely from them. The biological mechanisms are further supported by laboratory in vitro research and research with animals.

Several of these exposure-response functions (or curves) have been developed using data from European populations, for explicit use in Europe, and these are the ones we use at the European Environment Agency.

Specifically, deaths and disability-adjusted life years (a measure of the total burden of disease) from exposure to air pollution are calculated through comparative risk assessment methods. These estimate the possible reduction in disease if the exposure to air pollution was reduced to a minimum risk-level, assuming all other risk factors remain unchanged. These methods have been used and reviewed, and their results have been validated globally for about 25 years.

It is by applying these methods that we can estimate the share of deaths attributable to air pollution in Europe, even though death certificates do not directly specify it. However, it is important to be clear that this share of attributable deaths cannot, for the reasons explained above, be  linked to specific individual deaths. Instead, it reflects an estimate applied to the overall mortality of a population.

You can find out more about this methodology in this report.

We gratefully acknowledge the helpful inputs of Professors Thomas Muenzel (Department of Cardiology at the University Medical Centre at Johannes Gutenberg University in Mainz, Germany) and Jos Lelieveld (Max Planck Institute for Chemistry in Mainz, Germany).