Understanding global causes of death, GBD Learning

What are the leading causes of death in the world? This Global Burden of Disease (GBD) Learning video takes us through data and methods from the newly released GBD 2023 study—findings from a GBD 2023 capstone published in The Lancet titled, "Global burden of 292 causes of death in 204 countries and territories and 660 subnational locations, 1990–2023: a systematic analysis for the Global Burden of Disease Study 2023."

Video transcript

This transcript has been lightly edited for clarity

What are the most common causes of death in the world? And at what age are we dying from these causes?

We try to answer these questions in the newest report from the Global Burden of Disease, or GBD.

This report provides an analysis of 292 causes of death by age, sex, and year in 204 countries and territories between 1950 and 2023.

As with every GBD report, in GBD 2023, we reanalyzed the entire time series and estimated causes of death using CODEm, a modeling tool developed for GBD.

It all starts with data collection. We used more than 55,000 data sources and added more than 11,000 new data sources since GBD 2021. This includes data from vital registration systems, sample registration systems, verbal autopsies, surveys, censuses, and disease registries. Since the last GBD, we improved data coverage, reduced bias, and improved transparency in our methods.

Next, we clean and prepare the data for modeling. Causes are mapped to the 292 causes we estimate, and vague or misreported causes of death are redistributed to our causes systematically.

COVID-19 caused a major disruption to deaths globally, and many COVID-19 deaths were misclassified as a result. We needed to develop a special adjustment to capture missing cases of COVID-19.

We used a machine learning algorithm to scan for any unusual spikes in deaths from other diseases during the pandemic. We then determined whether each spike was due to COVID-19 or a true increase in cause-specific mortality and adjusted the estimates, reassigning a portion of that excess mortality to COVID-19.

Once data is prepared, we use CODEm, or the Cause of Death Ensemble model, to estimate death rates from all available data for each cause.

Once CODEm models from individual causes are vetted, we need to make sure all the deaths from every cause, for every location, year, age, and sex, add up to the all-cause mortality rates we estimate from demographic data. COVID-19 was estimated separately using a new model called OneMod. We then use another model called CoDCorrect to reconcile cause-specific deaths to add up to the total number of deaths.

After CoDCorrect, we add in deaths from unexpected events, such as natural disasters or human conflict.

Once we have the final cause-specific death rates, we are able to calculate a host of measures that help us better understand global death patterns. The key measures we focused on in this report are YLLs, probability of premature death, and mean age of death.

YLLs are “the product of the number of deaths for each cause-age-sex-location-year and the standard life expectancy at each age.” (GBD 2023 Causes of Death capstone)

We measured probability of premature death as the probability of dying before age 70 (70q0). We are able to measure premature death due to specific causes using this measure, advancing our understanding of the relationship between cause of death and age of death.

Mean age of death is “calculated by first assigning the midpoint age of each age group for every death, followed by computing the mean of all midpoint ages across all deaths attributed to a given cause.” (GBD 2023 Causes of Death capstone)

Additionally, we created expected age of death to compare against. “The expected mean age reflects the expected average age at death for individuals within a population, based on global mortality rates and the population’s age structure.” This allows us to compare a country’s performance against a global benchmark, gauging their progress toward global development goals. It compares a country’s actual age of death to an expected age based on global data, and that tells you if people in that country are dying from a disease earlier or later than they should be.

We also ran an analysis of the correlation of mean age of death and the Socio-demographic Index (SDI), a composite measure of fertility, educational attainment, and income per capita. This analysis assessed how SDI influences both the risk and timing of mortality.

Additionally, this report provides the most comprehensive analysis to date of the burden of COVID-19 deaths worldwide.

Finally, we added a few new causes compared to GBD 2021: ulcerative colitis, Crohn’s disease, thyroid diseases, “other endocrine, metabolic, blood and immune disorders,” and electrocution.

So, let’s look at the main findings from this study.

Cause-specific deaths: Now let’s look at cause-specific results. Ischemic heart disease and stroke are the top two global killers nearly every year, all the way back to 1990.

Since 1990, we have seen huge drops in death rates from conditions like diarrheal diseases, tuberculosis (TB), and measles. The age-standardized mortality rate, which we can use to compare death rates across different times and places, has plummeted for

these once-major killers. Diarrheal diseases, TB, and measles have dropped about 75% in some cases. COVID-19 explains the jump in “respiratory infections and TB” from 2020 to 2022.

In addition, there was a 66.5% decrease in YLLs from vaccine-preventable diseases between 1990 and 2023. Diarrhea, malaria, and tuberculosis, childhood killers, have also seen big drops in YLLs since 1990.

Meanwhile, we’re seeing increases in non-communicable diseases, with particularly large increases in diabetes.

COVID-19: This study also captures the single biggest shock to global health so far in the 21st century, the COVID-19 pandemic. We estimate that 18 million people died from COVID-19 between 2020 and 2023. By 2021, it was the number one cause of death in the entire world, replacing heart disease and stroke. But by 2023, it had fallen all the way down to 20th place.

Mean age of death: Back in 1990, the global all-cause mean age of death was just 46.8 years old. That leapt to 63.4 in 2023, an incredible 16.6 extra years of life on average in just over three decades.

In high-income countries, people are dying from drug use disorders at a much younger age than you’d expect. Meanwhile, in sub-Saharan Africa, women are dying from heart disease 3.4 years younger than expected.

Probability of premature death: Probability of premature death is the statistical chance that a person will die before they turn 70.

The GBD estimates significant decreases in lower respiratory infections, a nearly 50% drop in probability of premature death.

But we are moving backwards in other areas. The risk of a man dying before 70 from diabetes increased by over 75%. And for women, the risk of premature death from drug use disorders is up by 68%.

So, let’s nail down the main findings.

Number one, we’ve seen massive declines in mortality rates from diarrheal diseases, tuberculosis, stomach cancer, and measles, as well as considerable decreases in YLLs from those causes.

Two, the COVID-19 pandemic was seismic but temporary, becoming the leading cause of death in 2021 but dropping to 20th place by 2023.

And third, there are emerging challenges in chronic diseases, with increasing mortality rates for non-communicable causes like diabetes, as well as kidney disease, Alzheimer’s, and drug use disorders.

Fourth, the mean age of death is much higher than it was 30 years ago.

And finally, the probability of premature death (before age 70) has decreased in all super-regions, with substantial variation across locations.

The GBD 2023 report shows that global health is improving in some ways but facing new challenges.

While global progress has been substantial, significant health disparities persist between regions and populations, requiring targeted interventions and continued international cooperation. Pandemics and increasing non-communicable diseases both put strains on our existing health systems.

We need strong, resilient health systems that can handle the sudden shock of a new virus and manage the slow burn of NCDs at the same time. Providing strong, up-to-date estimates of causes of death helps inform those systems, ultimately improving global health equity.