It was updated in April 2019 with the latest statistics for the year 2017.
As global population increases, life expectancy rises, and living standards improve, causes of death across the world are changing. In this entry we present a global overview of the causes of death. The data visualizations and explainers in the early part of this entry attempt to provide a comparison between causes and risk factors of death across countries and age groups. The sections which then follow explore the empirical data of specific causes in more detail.
Each of the charts which follows is shown at a global level. However, all of the data included in this entry is available to explore at the national level using the “change country” function on the interactive charts. As will become clear in the data which follows, causes of death across the world remain heterogeneous and continue to change.
In the charts here we provide a comparison of the number of people who died as a result of specific causes in 2016. This data is given as the total across all ages, and both sexes. Causes of death varies notably between age groups — deaths across individual age groupings can be explored in the sections which follow.
At a global level we see that the majority of deaths are attributed to the category of non-communicable diseases (NCDs); these are chronic, long-term illnesses such as cardiovascular diseases (including stroke), respiratory disease, cancers and diabetes. Collectively NCDs account for more than 70 percent of global deaths.
Causes of death vary significantly by country and income levels across the world. These deaths are shown in the charts here, which present figures as absolute number of deaths, and as a percentage of total deaths. We also look at the major differences in mortality across the world using country examples in our blog post here. NCDs not only dominate mortality figures at a global level, but also account for the majority of deaths in high-income countries. Deaths from causes such as infectious disease, malnutrition, nutritional deficiencies, neonatal and maternal deaths in high-income countries is typically very low in relative terms.
These causes of death are, however, still prevalent (and some cases dominant) across low-to-middle income nations. In Kenya, for example, the leading cause of death remains diarrheal diseases. In South Africa and Botswana, the leading cause of death remains HIV/AIDS.
Death rates related to disease, illness and other health factors tend to change relatively slowly over time. Whilst death rates may fall or decline from year-to-year as part of a general trend, dramatic changes in such deaths are typically rare. Natural disaster and terrorism-related deaths are an important exception to this rule: mortality from disasters or terrorism/conflict events can be volatile. This can make the annual comparison of deaths and death rates between health-related factors and volatile events more challenging. Understanding the relative risk of these events can require a longer-term overview of high and low-mortality years. We cover discussion and analysis on this topic in a blog post here.
Whilst terrorism-related deaths are typically very low for most countries relative to other mortality causes, in some cases they can be significant. For example, terrorism fatalities were the 6th largest killer in Iraq in 2016. This is well above deaths related to road accidents, respiratory disease and dementia.
As we see in the charts above — whilst death rates are falling globally, the total number of deaths is increasing. What explains this increase in absolute number of deaths?
In the visualization — based on figures published by the Institute of Health Metrics and Evaluation (IHME) 1 — we see the global change in deaths and burden of disease over the past decade (2006 to 2016) disaggregated by contributing factors.
Population growth and ageing play a significant role in the total increase in number of deaths. If all other factors were kept constant, population growth alone over this period would have resulted in a 12.4 percent in deaths; population ageing alone would have resulted in a 14.9 percent increase. The impact of population ageing on non-communicable diseases (NCDs), the results are even more striking; alone, ageing would have resulted in a 19.5 percent increase in global deaths.
The actual net percentage change in deaths over this period has of course been smaller (approximately 3 percent) because population growth and ageing factors have been offset by reductions in risk exposure and residual factors (such as improved underlying health and healthcare). The impact of residual factors such as healthcare alone would have resulted in a 15.3 percent decline in global deaths. This influence has been even greater for communicable, maternal, neonatal and nutritional diseases — alone it would have resulted in a 30 percent decline in deaths.
In the chart we see the total annual number of deaths differentiated by broad age category. In 2017, we see there were around 56 million deaths globally; nearly half of these (49%) were in those aged 70 years or older; 27% in the 50-69 year age group; 14% aged 15-49; only 1% aged 5-14; and around 10% in children under the age of 5.
The demographics of deaths has changed significantly since 1990 with a notable shift from young-to-old. In 1990 nearly one-quarter of all deaths were in children under 5 years old. In 2017, this had declined to just under 10 percent. In contrast, the share of deaths in the over-70s age bracket has increased from 33 percent to 49 percent over this period.
The distribution of deaths by age varies significantly by country. In Japan, for example, around 83 percent of deaths occur in those aged 70 or older. Deaths in children and adolescents less than 15 years old account for around 0.25 percent of the total. Contrast this with South Sudan where nearly half (48 percent) occur in children under 5 years old. In South Africa, where the leading cause of death is HIV/AIDS, the leading age bracket is aged 15-49.
The chart shows the number of deaths in children under 5 years old by cause. Through the combination of neonatal (newborn infants less than 28 days old) disorders, infections and congenital (from birth) defects, we see that the largest share of deaths in under-5s arises from complications at birth or in the first few weeks of life. Under-5s are also highly susceptible to lower respiratory infections, infectious diseases, diarrheal infections, malnutrition and nutritional deficiencies.
Death rates in under-5s are typically much lower in high-income countries, and the nature of these deaths is different from lower incomes. In the United Kingdom, for example, child deaths tend to be highly dominated by neonatal complications; deaths from infectious and diarrheal diseases and malnutrition is very low. In contrast, infectious diseases and nutritional deficiencies are large causes of death in lower-income countries.
Globally, deaths in the 5-14 year old age bracket account for a small percentage of the total (1-2 percent). There are six dominant causes of deaths in this age category. The leading causes globally in 5-14 year olds are road accidents, cancers and malaria. Lower respiratory infections, HIV/AIDS, diarrheal diseases, and drowning are all dominant causes typically in the range of 40,000-50,000 deaths in 2017.
Again, this distribution varies by country. In the United States, for example, cancers are the leading cause of death. In India, it’s diarrheal diseases; in Bangladesh and China it’s drowning; and in South Africa HIV/AIDS.
By the 15-49 years old category, we see that non-communicable diseases (NCDs) begin to become dominant. Globally the leading cause of death in this age group is cardiovascular disease, followed cancers which both account for more than one million deaths. Road accidents, HIV/AIDS and suicide are all significant within this group.
For some countries, such as South Africa, by far the dominant cause of death (across all age groups) is HIV/AIDS in the 15-49 years old age bracket. In a number of countries (in particular across Latin America, including Brazil and Mexico), homicide is the dominant cause for 15-49 years old.
By aged 50-69 years old, non-communicable diseases (NCDs) are strongly dominant — here cardiovascular disease, cancers, respiratory disease and diabetes are the top causes. With the exception of HIV/AIDS and tuberculosis which for some countries climb into the top causes, the global variability in death causes for 50-69 year olds is much lower than that of younger age categories.
For the oldest age category (70 years and older), non-communicable diseases (NCDs) still dominate, however other death causes including Alzheimer’s/dementias, and diarrheal diseases also become dominant. Diarrheal diseases remain within the few leading causes of deaths in 70+ year olds for many low-income countries, despite being relatively low at higher incomes.
Whilst the Global Burden of Disease (GBD) assessment assigns each death to one specific cause, we know that the risk of disease burden and health outcomes are closely linked to a number of risk factors. It includes risk factors across four broad categories: behavioural, environmental, occupational, and metabolic risks.
We describe further information on how the GBD attributes risk factors to mortality in our Data Quality & Definitions section. These figures show the numbers of deaths in which each risk factor was implicated as a contributing cause; but because risk factors often occur together and interact, some deaths appear more than once. You should therefore not add up the deaths for each risk factor and expect the total to match the overall number of deaths.
High BMI, for example, may likely be present with other lifestyle factors such as low physical activity levels, high blood pressure, low fruit and vegetable intake. All of these estimates are developed independently. We cannot therefore sum all ‘attributed deaths’ and conclude that this is the actual number of deaths. The attributed number of deaths by risk factor in many cases exceeds that of those by cause of death.
In the chart we see the number of deaths attributed to specific risk factors in 2017. This data is measured across all age groups and both sexes — figures for specific age groups are detailed below. Here we see that there are several dominant risk factors for death: notably, those related to dietary and activity lifestyle factors (including blood pressure, physical activity, body-mass index, blood sugar, and dietary intake); smoking; air pollution (both outdoor and indoor); environmental factors including clean water and sanitation; and safe sex (for the prevention of HIV/AIDS).
The contribution of specific risk factors varies significantly by country. For most high-income countries, the dominant risk factors are those related to healthy diets, smoking and alcohol intake. Other risk factors such as clean water, sanitation, and child wasting or stunting are very low. In low-income countries the inverse is true: in Sierra Leone for example, the top risk factors include child wasting, household air pollution, unsafe water source, poor sanitation, and no access to handwashing facilities. For countries where HIV/AIDS is a major health burden, such as South Africa and Kenya, unsafe sex is the top risk factor.
In the charts here we see how these risk factors differ between age groups, ranging from children under-5, to those 70+ years old. In the youngest age group we see dominant risk factors related to childhood development, including child wasting, stunting and low-birth weight. In older children and adolescents, environmental factors including water sources, sanitation, handwashing and air pollution become more dominant. By the 15-49 year old category, unsafe sex becomes the largest risk factor (in relation to the contraction of HIV/AIDS), with alcohol, smoking and some dietary factors also becoming important.
By the 50-69 year old and 70+ year old age category, lifestyle factors such as dietary intake, smoking, alcohol, and low physical intake become dominant — ultimately increasing the risk of non-communicable diseases (NCDs) such as cardiovascular diseases, cancers, stroke, respiratory disease and diabetes.
Causes of death over the long-run are harder to assess due to inconsistencies and poor coverage of data reporting. Extensive data is available for few countries. In the charts here we see causes of deaths across the 20th century in the United States, as reported by the Center for Diseases Control (CDC). Even in this case, data coverage is patchy and only available for the top ten causes of death per year. Data omissions may therefore result when particular causes of death fall in and out of this top ten listing.
Overall, over the 20th century we see a notable transition in deaths from communicable and infectious diseases such as tuberculosis, pneumonia and influenza towards non-communicable diseases (NCDs) such as heart disease, cancers, diabetes and respiratory disease. This is a similar transition to what we see across many developing countries today.
Cardiovascular disease (CVD) is a term used to refer to the range of diseases which affect the heart and blood vessels. These include hypertension (high blood pressure); coronary heart disease (heart attack); cerebrovascular disease (stroke); heart failure; and other heart diseases. Cardiovascular disease data includes figures related to strokes, however we also cover this separately in the section below. Cardiovascular disease is the top cause of death globally.
In the visualization we see the breakdown of deaths from CVD by age category. Globally we see that approximately 17.8 million people died from CVD. The majority (64 percent) of deaths occurred in the age bracket of 70 years and above. Just below 30 percent were aged 50-69, and the remaining were aged 15-49 (CVD deaths in those aged 14 years and under are small).
In the visualization we see cardiovascular disease death rates, expressed as the number of CVD deaths per year per 100,000 individuals. Note that these rates have been age-standardized which aims to correct for differences in the age structure of a population (which are different between countries and change over time). This therefore allows us to compare the likelihood that any given individual will die from cardiovascular disease across countries and through time.
Overall we see a strong East-West divide in CVD death rates. Rates across North America and Western/Northern Europe tend to be significantly lower than those across Eastern Europe, Asia and Africa. Across most of Latin America, these rates are moderate. In France, for example, the age-standardized rate was around 86 per 100,000 in 2017; across Eastern Europe this rate was around 5 times higher at 400-500 per 100,000. At the highest end of the scale, Uzbekistan had a rate of 724 per 100,000.
In the chart we see the CVD death rate per 100,000 differentiated by age categories. As established in the breakdown of CVD deaths by age, the majority of deaths occur in the 70+ years old age category. Death rates are therefore significantly higher in the oldest age group at over 2600 per 100,000 in 2016.
Cancers are defined by the National Cancer Institute (NCI) as a collection of diseases in which abnormal cells can divide and spread to nearby tissue. As this definition suggests, cancers can arise in many parts of the body (leading to a range of cancer types, as shown below) and in some cases spread to other parts of the body through blood and lymph systems.
In the chart we see annual deaths attributed to cancers, differentiated by type. Globally, around 8.9 million deaths were attributed to cancers in 2016 — an increase from around 5.7 million in 1990. You can also view cancer deaths in specific countries using the “change country” function.
As shown, there are a broad range of cancer types with varying magnitudes of total death. At a global level, tracheal and lung cancer is the largest killer with over 1.7 million in 2016, followed by stomach, colon and rectum, and liver cancer. On relative terms, testicular, thyroid and non-melenoma skin have the lowest death toll of cancers at the global level.
In the chart we see total cancer deaths broken down by age grouping. As we see, the majority of deaths arise in those over 50 years old, with 46 percent over 70 years old and 41 percent between 50-69 years old. Cancer deaths in children are much smaller in comparison (less than one percent of the total are in children under 14 years old) and typically arise as leukemia cases.
In the chart we see cancer death rates, expressed as the number of cancer deaths per year per 100,000 individuals. Note that these rates have been age-standardized which aims to correct for differences in the age structure of a population (which are different between countries and change over time). This therefore allows us to compare the likelihood that any given individual will die from cancer across countries and through time.
Cancer death rates vary across the world, but for most death rates are below 150 per 100,000 individuals.
If we use the timeline to look at how cancer rates have changed with time we see that for many countries, this rate has declined since 1990. For example, rates in the United States have fallen from 160 to 128 per 100,000; in China from 163 to 138 per 100,000; in the United Kingdom from 179 to 139 per 100,000.
In the visualization we see these age-standardized death rates (measured per 100,000 people) for individual cancer types. Declines in cancer rates vary by type: for many, such as prostrate and kidney, progress has been relatively slow. For some, such as stomach cancer there has been notable progress — declining from 19 to 11 deaths per 100,000.
In the visualization we see how cancer death rates vary by age group. As we see, death rates are highest for those aged over 70 years old with nearly 1,000 deaths per 100,000 people. Rates in the 50-69 years old category are more than three times lower at just under 300 per 100,000 individuals. For children aged below 14 years old, rates at a global level are less than 5 per 100,000.
Chronic respiratory diseases (CRDs) are diseases which affect the airways and overall structure of the lungs. Common CRDs include chronic obstructive pulmonary disease (COPD), asthma, occupational lung diseases and pulmonary hypertension.
The chart shows the breakdown of deaths related to chronic respiratory diseases (CRDs) by age group. Globally in 2017, around 3.9 million died from CRDs. More than two-thirds (69 percent) of respiratory disease deaths were from those aged 70 years and older, with a further 26 percent in the 50-69 year group; and the remaining 4-5 percent aged 49 years and younger.
In the visualization we see respiratory disease death rates, expressed as the number of CRDs deaths per year per 100,000 individuals. Note that these rates have been age-standardized which aims to correct for differences in the age structure of a population (which are different between countries and change over time). This therefore allows us to compare the likelihood that any given individual will die from CRDs across countries and through time.
In most countries across the world, the age-standardized death rate from CRDs are below 40 per 100,000 people. Rates across Sub-Saharan Africa and South-East Asia are typically slightly higher (between 40 and 60 per 100,000), and are typically highest across South Asia (over 130 per 100,000 in India and Nepal). At a country-level the highest rate is in Papua New Guinea at 351 per 100,000.
Respiratory disease death rates have fallen dramatically in some countries over the past few decades. Since 1990, death rates in China have fallen around 70 percent from 200 to 63 per 100,000. In India, rates have fallen by around 40 percent since 1990.
In the visualization we see the breakdown of respiratory disease death rates by age group. Here we see that death rates in the oldest age category (70+ years old) are far above those of other age groups.
The World Health Organization (WHO) defines a stroke as the “interruption of the blood supply to the brain, usually because a blood vessel bursts or is blocked by a clot. This cuts off the supply of oxygen and nutrients, causing damage to the brain tissue.” Strokes are included as a sub-category within cardiovascular diseases.
In the chart we see the breakdown of total deaths attributed to strokes by age group. Globally it’s estimated that 6.2 million people died from a stroke in 2017. The majority (64 percent) of deaths from stroke arise in those aged 70 years or older. Those aged 50-69 years old account for approximately 30% of deaths, with those aged 49 and under collectively account for around 6 percent.
In the visualization we see stroke death rates, expressed as the number of stroke deaths per year per 100,000 individuals. Note that these rates have been age-standardized which aims to correct for differences in the age structure of a population (which are different between countries and change over time). This therefore allows us to compare the likelihood that any given individual will die from stroke across countries and through time.
Stroke death rates vary significantly across the world, ranging from less than 25 to over 200 per 100,000.
As we see in the breakdown of stroke deaths by age, we see that those over the age of 70 have the highest risk of death at over 900 deaths per 100,000 in 2017. This is shown, relative to other age categories in the chart.
Dementia comprises several forms — the most common being Alzheimer’s disease — is an illness which results in a deterioration of cognitive capacity and function beyond what is expect from the normal ageing process. It can occur either in a chronic or progressive form. It affects several cognitive functions including memory, comprehension, judgement, language and learning capacity.
The chart shows the breakdown of dementia-related deaths by age group. Dementia typically occurs in older persons: of the 2.4 million who died from dementia in 2017, 94 percent were 70 years or older.
The visualization shows dementia death rates, expressed as the number of dementia deaths per year per 100,000 individuals. Note that these rates have been age-standardized which aims to correct for differences in the age structure of a population (which are different between countries and change over time). This therefore allows us to compare the likelihood that any given individual will die from dementia across countries and through time.
Across most countries, the death rate from dementia-related illness is below 55 per 100,000 individuals. Dementia rates in some countries have changed slightly since 1990, but significantly less so than other disease burdens.
Diarrheal disease remains one of the largest causes of death in children, and in some countries (e.g. Kenya), it is the top mortality cause. Diarrheal diseases are typically a symptom of infections within the intestinal tract and is contracted through poor hygiene, sanitation, unsafe water sources, or contaminated food. Most deaths result from dehydration and the loss of essential nutrients and salts. Diarrheal disease is both preventable and treatable.
You can find a more in depth look at the burden of diarrheal diseases and how we can prevent them in our enry on the topic here.
The chart shows the breakdown of diarrheal disease deaths by age group. Overall we see there has been a significant reduction in global deaths since 1990: falling from 2.6 million to 1.6 million in 2017. Diarrheal deaths disproportionately affect the young and the old: 40 percent of global deaths were from those aged 70 years or older, with one-third under 5 years old.
Since 1990 there has been an even more dramatic reduction in deaths in children under 5 years old: in 1990 they accounted for 65 percent of deaths, which have fallen as a share by around half to 34. In absolute terms, this is a reduction from 1.7 million to 534,000 (a reduction of more than two-thirds).
In the chart we see diarrheal disease death rates, expressed as the number of deaths per year per 100,000 individuals. Note that these rates have been age-standardized which aims to correct for differences in the age structure of a population (which are different between countries and change over time). This therefore allows us to compare the likelihood that any given individual will die from diarrheal disease across countries and through time.
Overall, we see that the highest death rates lie across Sub-Saharan Africa and South Asia, where rates can reach over 150 per 100,000 (as in the case of the Central African Republic). Rates across these regions typically range from 50 to 150 per 100,000. Across most of the rest of the world, rates are below 25 per 100,000 and in many cases below 1 per 100,000.
In the chart we see how diarrheal disease death rates vary by age category. As with the total number of deaths by age, we see the highest death rate in 70+ year olds. At a global level, death rates of under-5s are almost 50 percent lower, and the remaining age categories far below these rates.
Malnutrition arises in various forms, with the broad definition capturing undernourishment, micronutrient deficiencies and obesity. In this case, we refer to ‘protein-energy malnutrition‘ (PEM) which refers to energy or protein deficiency caused by insufficient food intake. Protein-energy deficiency can also be exacerbated by infection or disease, which can have the effect of increasing nutritional needs, and/or reducing the body’s ability to retain energy or nutrients. You can find more information on hunger and undernourishment in our entry.
In the chart we see the annual number of deaths attributed to protein-energy malnutrition (PEM), differentiated by age group. Globally there were approximately 232,000 deaths related to PEM. Children under 5 years old are disproportionately affected by PEM (accounting for 61 percent of global deaths) — child wasting (too little weight for one’s height) and stunting (too short for one’s age) is a common symptom of malnutrition.
In the chart we see death rates from protein-energy malnutrition, expressed as the number of deaths per year per 100,000 individuals. Note that these rates have been age-standardized which aims to correct for differences in the age structure of a population (which are different between countries and change over time). This therefore allows us to compare the likelihood that any given individual will die from PEM across countries and through time.
The highest rates are seen across across Sub-Saharan Africa, which are typically in the range of 10-100 per 100,000 individuals. For most countries, this rate is below 5 per 100,000. In North Korea during its famine period, rates reached over 400 per 100,000.
In the visualization we see the breakdown of death rates by age category. As with the total number of deaths by age, rates in children under 5 years old are highest; at a global level, these have fallen by around two-thirds from 63 to 20 per 100,000 since 1990. Rates for those over 70 years old are also relatively high, at 12 per 100,000 (although this decline over time has been less significant).
Tuberculosis (TB) is an illness caused by the ingestion of bacteria (Mycobacterium tuberculosis) which affects the lungs. The World Health Organization (WHO) estimate that up to one-quarter of the global population has latent TB, meaning they have been infected with the disease but are not ill with the disease (although this does not inhibit it from becoming active in the future).
People with compromised immune systems, such as those suffering from malnutrition, diabetes, or are smokers are more likely to become ill with TB. There is a strong link between HIV/AIDS and TB: those infected with HIV are 20-30 times more likely to develop active tuberculosis.
In the chart we see the breakdown of deaths from tuberculosis by age category. At a global level there has been a 25% drop in TB deaths, falling from 1.6 million in 1990 to 1.2 million in 2017.
TB deaths are relatively equally distributed across age categories over 15 years old: around 31% occur in those aged 15-49; 36% in those aged 50-69 years old; and 26 percent in those aged 70+ years old.
In the chart we see death rates from tuberculosis, expressed as the number of deaths per year per 100,000 individuals. Note that these rates have been age-standardized which aims to correct for differences in the age structure of a population (which are different between countries and change over time). This therefore allows us to compare the likelihood that any given individual will die from tuberculosis across countries and through time.
Across most countries, the death rate from TB is below 5 per 100,000. Rates in 2017 across Eastern Europe were slightly higher, between 5-10 per 100,000. Across South Asia, these reach 25-50 per 100,000, with highest rates across Sub-Saharan Africa ranging from 50 to over 250 per 100,000.
In the visualization we see the breakdown of death rates from TB by age category. Here we see that death rates from TB are highest in the 70+ years old age category, followed by 50-69 year olds.
We cover the topic of HIV/AIDS deaths, breakdown by age, death rates, and additional relevant metrics in our full entry on HIV/AIDS.
We cover the topic of malaria deaths, breakdown by age, death rates, and additional relevant metrics in our full entry on Malaria.
The World Health Organization (WHO) emphasises that drowning is one of the most overlooked, preventable causes of death across the world.3 For every country in the world, drowning is among the top 10 killers for children. In some countries, such as Bangladesh, it is the top mortality cause for children under 15 years old.
In the chart we see the breakdown of annual drowning deaths by age group. At a global level we see drowning deaths have declined around 45 percent from 530,000 in 1990 to 300,000 in 2017. The most dramatic decline is seen in under-5s, where deaths have fallen from 209,000 to just over 60,000 over this period.
In the chart we see death rates from drowning, expressed as the number of deaths per year per 100,000 individuals. Note that these rates have been age-standardized which aims to correct for differences in the age structure of a population (which are different between countries and change over time). This therefore allows us to compare the likelihood that any given individual will die from drowning across countries and through time.
In 2016, death rates were highest in Papua New Guinea and Seychelles, between 10 to 16 deaths per 100,000. Rates were also high in countries such as Bangladesh, Central African Republic, Vietnam, and Haiti.
If we look at death rates we see a significant decline since 1990 — especially in low to middle-income countries. In Bangladesh and China, for example, rates have fallen by more than two-thirds over this period.
In the visualization we see the relative death rates from drowning across age groups. Here we see that both the young (under 5s) and old (70+ years old) are at highest risk of death from drowning. Most striking is the dramatic decline in death rates for under-5s. This has fallen by more than two-thirds since 1990, decreasing from 32 to 9 per 100,000.
The chart shows the annual deaths from fire or burning incidents broken down by age group. In 2017, there were around 121,000 global deaths from fire which represents a slight decline from the mid-1990s when deaths reached over 150,000.
In the chart we see death rates from fire, expressed as the number of deaths per year per 100,000 individuals. Note that these rates have been age-standardized which aims to correct for differences in the age structure of a population (which are different between countries and change over time). This therefore allows us to compare the likelihood that any given individual will die from fire across countries and through time.
Most countries across the Americas, Western Europe, East Asia and Oceania average death rates below 2 per 100,000. Rates across other regions are typically higher at 2-6 per 100,000. When viewed through time we see a notable decline in fire death rates, particularly across Sub-Saharan Africa and Eastern Europe.
In the visualization we see the relative death rates between age categories. At the global level, those 70 years and above are typically at the highest risk with 7 per 100,000. Next is those under 5 years old, but with a significant drop in death rate to 2-3 per 100,000.
Road incident deaths include those from motor vehicles (including drivers and passengers within the car), pedestrians, motorcyclists and cyclists. In the chart we see the breakdown of these deaths by category. The total number of road deaths increased during the 1990s and early 2000s before roughly plateauing since then (with a small decline in recent years).
The largest share of deaths at the global level are pedestrians (with 39 percent of the share), closely followed by those in motor vehicles (36 percent); motorcyclists (18 percent); and cyclists (6 percent). This share has remained fairly consistent with time.
This breakdown varies significantly across the world, however. In India and Brazil, for example, a much larger share (31 and 33 percent respectively) are motorcyclists. In the United States 68 percent are motor vehicle passengers or drivers. In China, 58 percent are pedestrians.
In the chart we see long-term trends in the number of deaths from road incidents. Note that these estimates may differ from IHME estimates above; most national estimates are based solely on police reports; IHME attempt to correct for varying factors in poor or under-reporting of road deaths. We discuss this with comparisons in our section on Data Quality and Definitions.
As we see from the two charts here: road deaths have been on the decline over the long-term. However, this is not attributed to a reduction in road accidents; in fact, in Germany the number of reported incidents has increased substantially over this period. Whilst some of this may be attributed to improved and better reporting, it is also true that the numbers injured in road accidents has not declined. Overall, we may therefore conclude that safety standards of and within cars, or in the severity of road incidents due to interventions such as speed limits have dramatically decreased the mortality risk of a given accident.
In the chart we see the breakdown of road accident deaths by age category. The largest share of deaths at a global level are within the 15-49 year old category with 54% of the total.
In the chart we see death rates from road incidents, expressed as the number of deaths per year per 100,000 individuals. Note that these rates have been age-standardized which aims to correct for differences in the age structure of a population (which are different between countries and change over time). This therefore allows us to compare the likelihood that any given individual will die from road incidents across countries and through time.
Death rates are typically lowest across Western Europe and Japan, with less than 5 deaths per 100,000 individuals. Across the Americas, rates are typically slightly higher at 5 to 20; most countries in Asia lie between 15 and 30; and rates are typically highest across Sub-Saharan Africa with over 25 per 100,000.
In the chart we see the breakdown of death rates from road incidents by age category.
We cover the topic of natural disasters, number of deaths, death rates, and additional relevant metrics by disaster type in our full entry on Natural Disasters.
Amnesty International attempt to quantify the number of executions (carried out by the state) across the world. These figures are shown in the chart from 2007 onwards. Note that these represent the number of executions carried out, and not the number sentenced to death or on death row (which are much higher).
In 2016, at least 1,032 people were executed across 23 countries. This figure is given as a minimum because data collection in some countries is challenging to assess. In 2009 Amnesty International stopped attempts to accurately quantify Chinese figures. They do however maintain that China has the highest annual figures, and reports a minimum of 1000 deaths per year.
We cover the topic of suicide deaths, breakdown by age, death rates, and additional relevant metrics in our full entry on Suicide.
We cover the topic of smoking prevalence, deaths, death rates, secondhand smoke deaths, and additional relevant metrics in our full entry on Smoking.
We cover the topic of homicide deaths, breakdown by age, death rates, historical prevalence of violence and additional relevant metrics in our full entry on Homicides.
We cover the topic of terrorism, including the prevalence of incidents, deaths, injuries and additional relevant metrics in our full entry on Terrorism.
And close to a million people are killed by other animals in one year. The breakdown in the chart counts 913,000 deaths from 19 different species.4
Estimated number of global human deaths by animal, either from direct contact/attack or transmission of disease.5
One of the primary motivations for our work at Our World in Data is to provide a fact-based overview of the world we live in — a perspective that includes the persistent and long-term changes that run as a backdrop to our daily lives. We aim to provide the complement to the fast-paced reporting we see in the news. The media provides a near-instantaneous snapshot of single events; events that are, in most cases, negative. The persistent, large-scale trends of progress never make the headlines.
But is there evidence that such a disconnect exists between what we see in the news and what is reality for most us?
One study attempted to look at this from the perspective of what we die from: is what we actually die from reflected in the media coverage these topics receive?6
To answer this, Shen and his team compared four key sources of data:
- the causes of deaths in the USA (statistics published by the CDC’s WONDER public health database)
- Google search trends for causes of deaths (sourced from Google Trends)
- mentions of causes of deaths in the New York Times (sourced from the NYT article database)
- mentions of causes of deaths in The Guardian newspaper (sourced from The Guardian article database)
For each source the authors calculated the relative share of deaths, share of Google searches, and share of media coverage. They restricted the considered causes to the top 10 causes of death in the US and additionally included terrorism, homicide, and drug overdoses. This allows for us to compare the relative representation across different sources.7
So, what do the results look like? In the chart below I present the comparison.
The first column represents each cause’s share of US deaths; the second the share of Google searches each receives; third, the relative article mentions in the New York Times; and finally article mentions in The Guardian.
The coverage in both newspapers here is strikingly similar. And the discrepancy between what we die actually from and what we get informed of in the media is what stands out:
- around one-third of the considered causes of deaths resulted from heart disease, yet this cause of death receives only 2-3 percent of Google searches and media coverage;
- just under one-third of the deaths came from cancer; we actually google cancer a lot (37 percent of searches) and it is a popular entry here on our site; but it receives only 13-14 percent of media coverage;
- we searched for road incidents more frequently than their share of deaths, however, they receive much less attention in the news;
- when it comes to deaths from strokes, Google searches and media coverage are surprisingly balanced;
- the largest discrepancies concern violent forms of death: suicide, homicide and terrorism. All three receive much more relative attention in Google searches and media coverage than their relative share of deaths. When it comes to the media coverage on causes of death, violent deaths account for more than two-thirds of coverage in the New York Times and The Guardian but account for less than 3 percent of the total deaths in the US.
What’s interesting is that Americans search on Google is a much closer reflection of what kills us than what is presented in the media. One way to think about it is that media outlets may produce content that they think readers are most interested in, but this is not necessarily reflected in our preferences when we look for information ourselves.
[Clicking on the visualization will open it in higher resolution; The chart shows the summary for the year 2016, but interactive charts for all available years is available at the end of this blog.8]
As we can see clearly from the chart above, there is a disconnect between what we die from, and how much coverage these causes get in the media. Another way to summarize this discrepancy is to calculate how over- or underrepresented each cause is in the media. To do this, we simply calculate the ratio between the share of deaths and share of media coverage for each cause.
In the chart below we see how over- or underrepresented each cause is in newspaper coverage.9 Causes shown in red are overrepresented in the media; those in blue are underrepresented. Numbers denote the factor by which they are misrepresented.
The major standout here – I had to break the scale on the y-axis since it’s several orders of magnitude higher than everything else – is terrorism: it is overrepresented in the news by almost a factor of 4000.
Homicides are also very overrepresented in the news, by a factor of 31. The most underrepresented in the media are kidney disease (11-fold), heart disease (10-fold), and, perhaps surprisingly, drug overdoses (7-fold). Stroke and diabetes are the two causes most accurately represented.
[Clicking on the visualization will open it in higher resolution].
From the comparisons above, it’s clear that the news doesn’t reflect what we die from. But there is another important question: should these be representative?
There are several reasons we would, or should, expect that what we read online, and what is covered in the media wouldn’t correspond with what we actually die from.
The first is that we would expect there to be some preventative aspect to information we access. There’s a strong argument that things we search for and gain information on encourages us to take action which prevents a further death. There are several examples where I can imagine this to be true. People who are concerned about cancer may search online for guidance on symptoms and be convinced to see their doctor. Some people with suicidal thoughts may seek help and support online which later results in an averted death from suicide. We’d therefore expect that both intended or unintended exposure to information on particular topics could prevent deaths from a given cause. Some imbalance in the relative proportions therefore makes sense. But clearly there is some bias in our concerns: most people die from heart disease (hence it should be something that concerns us) yet only a small minority seek [possibly preventative] information online.
Second, this study focused on what people in the USA die from, not what people across the world die from. Is media coverage more representative of global deaths? Not really. In another blog post, ‘What does the world die from?‘, I looked in detail at the ranking of causes of death globally and by country. The relative ranking of deaths in the USA is reflective of the global average: most people die from heart disease and cancers, and terrorism ranks last or second last (alongside natural disasters). Terrorism accounted for 0.06 percent of global deaths in 2016. Whilst we’d expect non-US events to feature in the New York Times, global news shouldn’t substantially affect representative coverage of causes.
The third relates to the very nature of news: it focuses on events and stories. Whilst I am often critical of the messages and narratives portrayed in the media, I have some sympathy for what they choose to cover. Reporting has become increasingly fast-paced. As news consumers, our expectations have quickly shifted from daily, to hourly, down to minute-by-minute updates of what’s happening in the world. Combine this with our attraction to stories and narratives. It’s not surprising that the media focuses on reports of single (inadvertently negative) events: a murder case or a terrorist attack. The most underrepresented cause of death in the media was kidney disease. But with an audience that expects a minute-by-minute feed of coverage, how much can possibly be said about kidney disease? Without conquering our compulsion for the latest unusual story, we cannot expect this representation to be perfectly balanced.
Media and its consumers are stuck in a reinforcing cycle. The news reports on breaking events, which are often based around a compelling story. Consumers want to know what’s going on in the world — we are quickly immersed by the latest headline. We come to expect news updates with increasing frequency, and media channels have clear incentives to deliver. This locks us into a cycle of expectation and coverage with a strong bias for outlier events. Most of us are left with a skewed perception of the world; we think the world is much worse than it is.10
The responsibility in breaking this cycle lies with both media producers and consumers. Will we ever stop reporting and reading the latest news? Unlikely. But we can all be more conscious of how we let this news shape our understanding of the world.
As media consumers we can be much more aware of the fact that relying on the 24/7 news coverage alone is wholly insufficient for understanding the state of the world. This requires us to check our (often unconscious) bias for single narratives and seek out sources that provide a fact-based perspective on the world.
This antidote to the news is what we try to provide at Our World in Data. It should be accessible for everyone, which is why our work is completely open-access. Whether you are a media producer or consumer, feel free to take and use anything you find here.
Shares of deaths, media coverage and Google searches over time
The interactive charts present the full annual data series published by Shen et al. (2018), and summarised above, on the relative share of deaths in the US, Google searches and New York Times (NYT) and The Guardian media coverage across the 13 cause of death categories. For share of deaths, NYT and The Guardian coverage, data extends from 1999 to 2016. Due to data availability Google Trends data only runs from the year 2004 to 2016.
The IHME’s Global Burden of Disease (GBD) has developed a highly standardised approach to the attribution of deaths to specific causes based on data limitations, including incompleteness of vital registration (VR) data, the use of verbal autopsy (VA) studies in locations with incomplete VR records, the allocation of disease burden to categories which can be directly attributed to mortality causes, and general data insufficiency in some countries.12
The GBD assessment is strongly tied to mortality cause categories as defined within the International Classification of Diseases (ICD) codes, as used by the World Health Organization (WHO). The GBD methodology states that “each death is attributed to a single underlying cause — the cause that initiated the series of events leading to death—in accordance with ICD principles”. The GBD has developed a comprehensive and in-depth database with regards to mortality and attributed causes, utilising data from various sources. These sources include vital registration (VR); verbal autopsy (VA); surveillance, census and survey data; cancer registries; and police records. GBD then develops a data standardisation and processing methodology within which they define data quality and completeness scores, and where necessary adjust completeness to 100% using cause fractions for a given location-age-sex-year and estimated all-cause mortality for that location-age-sex-year.
An important step in the GBD methodology standardisation is in reallocating deaths attributed within ICD classifications without an underlying cause of death (for example, senility) which can be intermediate but not final cause of death. These categories are termed as ‘garbage codes’. GBD redistribute these garbage codes using a methodology explained in detail in Naghavi et al. (2010).13
Note that this redistributing of ‘garbage codes’ in some cases explains the difference in estimates between IHME and WHO, such as for road accident deaths (compared below).
Death and death rates analyses are then carried out by GBD across locations, age, sex and year based on its Cause of Death Ensemble model (CODEm). Full description of GBD methodology can be found here.
Estimating and attributing relative risk factors to number of deaths is a complex task — particularly when risk factors can compound and collectively influence the likelihood of disease burden. The IHME’s Global Burden of Disease (GBD) studies provide one of, if not the, most in-depth analysis and synthesis of relative risk factors.14
IHME differentiate risk factors into four broad categories: behavioural, environmental, occupational, and metabolic risks.
IHME have developed a Comparative Risk Assessment (CRA) conceptual framework15 by which they have built a web of risk factors or causes which affect health outcomes. For example, there is evidence of links between body mass index (BMI) and the risk of multiple non-communicable diseases (NCDs) including cardiovascular disease, ischemic stroke and some cancers.16
Such risk-outcome pairs (e.g. high BMI and ischemic stroke) are formed based on evidence links using methods such as cohort studies, randomised trials, and case-control studies. Once a risk-outcome pair has been formed, how does IHME begin to quantify the disease burden or number of deaths attributed to each risk? The CRA can be used for two different types of assessment: attributable burden and avoidable burden. ‘Attributable burden’ represents the reduction in current disease burden (or that of a given year) if population exposure had shifted to another counterfactual/hypothetical exposure level; ‘avoidable burden’ represents the potential future avoided burden if population exposure was to shift to a counterfactual level of exposure. Since the number of deaths is based on current or historical data, the data presented here is that of attributable burden.
Attributable burden effectively asks the question: “what would the number of deaths from ischemic stroke be if everyone’s body-mass index (BMI) was reduced to a healthy/optimal level (within range)?” This healthy/optimal level is defined as the ‘theoretical minimum risk exposure level (TMREL)‘. Cohort, case studies and trials of established risk-exposure relationships between BMI and ischemic stroke allow for the calculation of the reduction in deaths which would have occurred if BMI was reduced to a healthy level across the population distribution. This relationship can be established by specific demographic groups, such as by sex or age. The difference between the number of deaths from ischemic stroke which would have occurred at the TMREL and at the actual BMI distribution is given as the number of deaths attributed to high BMI from ischemic stroke.
By completing this process for all risk-outcome pairs, IHME can sum to estimate the total number of deaths attributed to high BMI, and replicated for all risk factors using their individual risk-outcome exposure curves. Note that this process of estimation is not additive; in other words, these risk-specific relationships do not account for the compounding effects of multiple risk factors. High BMI, for example, may likely be present with other lifestyle factors such as low physical activity levels, high blood pressure, low fruit and vegetable intake. All of these estimates are developed independently. We cannot therefore sum all ‘attributed deaths’ and conclude that this is the actual number of deaths. The attributed number of deaths by risk factor in many cases exceeds that of those by cause of death.
Full methodological explanation of the IHME’s approach to risk factor attribution can be found here.
The data presented on road mortality in this entry were based on figures published by the Institute for Health Metrics and Evaluation (IHME) Global Burden of Disease (GBD) programme. It has been widely recognised that IHME data on road accidents tends to be higher than that of official police reports and that reported by the World Health Organization — this tends to be the case for OECD countries.17 In the chart we compare road mortality figures by the IHME (y-axis) and World Health Organization (WHO)18 (x-axis). Also shown is the line of parity, along which IHME and WHO estimates are the same.
As we see, there are in some cases notable differences between these estimates, and in particular higher figures for OECD countries for IHME figures. Why does this discrepancy occur? In its own assessment report, IHME highlights this difference, noting that its estimates for OECD countries are on average 58 percent higher than official statistics.19
IHME note that official statistics ” typically relies on police reporting, while GBD relies primarily on death registration data in these countries. Furthermore, IRTAD only includes those deaths that occur within the first 30 days of a crash and excludes motor vehicle-related deaths that do not occur in traffic. In contrast, our estimates provide a more comprehensive estimate of the impact of motor vehicles on population health.”
A notable difference between these sources is the period within which deaths are reported. Official statistics only record that which occurs within 30 days of a crash. GBD data attempts to account for intermediate injuries of health burden which may result in death following a 30 day period. This relates strongly to discussion in our section on ‘estimating causes of death’ where so-called ‘garbage codes’ (which are ICD classifications which can be intermediate but not final cause of death) are assigned to road injuries. In this example, it may be extreme back pain. GBD attempts to redistribute these ‘garbage codes’ to take a more comprehensive estimate of health burden and death which results over a period longer than 30 days.
Institute of Health Metrics and Evaluation (IHME), Global Burden of Disease (GBD)
- Data: Death rates, absolute number of premature deaths and DALYS across all risk factors and causes
- Geographical coverage:Global, across all regions and countries
- Time span:Most metrics available from 1990 onwards
- Available at: Online here
World Health Organization (WHO) Global Health Observatory (GHO)
- Data: Causes-specific mortality by age and sex
- Geographical coverage:Global, by region and by country
- Time span:Most metrics available from 2000 to 2015 in 5-year incrememnts
- Available at: Online here
Global Terrorism Database (GTD)
- Data: Terrorist attacks with 45-120 variables for each, including number of fatalities, injuries, weapons used, and perpetrators
- Geographical coverage: Global by country
- Time span: 1970-2016
- Available at: http://www.start.umd.edu/gtd/
- Data: International reports of executions
- Geographical coverage: Global by country
- Time span: 2007-2016
- Available at: https://www.amnesty.org/en/what-we-do/death-penalty/