October 9, 2026 interva.net Editorial

Africa's Mortality Is Turning Chronic, Faster Than Its Death Records Can Follow

A GBD 2023 analysis counts 4.7 million non-communicable disease deaths in Africa in a single year — a burden whose estimate leans on verbal autopsy precisely because so many of these deaths are never medically certified.

verbal autopsyNCDscause of deathglobal health

In 2023, non-communicable diseases killed an estimated 4.70 million people in Africa. The figure comes from a systematic analysis for the Global Burden of Disease Study 2023, published in The Lancet Global Health on 21 September, and it carries an uncertainty interval of 4.47 to 4.93 million deaths. Ischaemic heart disease alone accounts for roughly 801,000 of them; stroke accounts for about 700,000.

The number is large enough that it invites a mistaken reading. Africa’s mortality profile has not simply swapped infections for chronic disease. What the analysis describes is a population that is both growing and ageing into a different mix of causes — and a measurement system that has to keep up with the change.

A burden that rises in absolute terms while its rates fall

The study’s central paradox is arithmetic. Between 1990 and 2023 the age-standardised NCD mortality rate in Africa fell by 9.0% (2.5–15.5), the incidence rate by 2.7% and the DALY rate by 6.4%. Yet absolute deaths, cases and disability-adjusted life-years all rose sharply — 2.33 billion incident cases and 263 million DALYs in 2023 — because population growth and demographic change outran the modest improvement in rates.

That is what an epidemiological transition looks like when it is measured honestly: better per-person risk on average, a much larger number of people at risk. The authors describe the pattern as a “paradox driven predominantly by rapid population growth and demographic changes” rather than by worsening age-specific disease.

The five causes on top, and the risks behind them

For all sexes and ages, the leading Level 3 causes of NCD death in Africa in 2023 were ischaemic heart disease (801,000), stroke (700,000), chronic kidney disease (293,000), hypertensive heart disease (246,000) and diabetes (245,000). Mental disorders (30.9 million DALYs) and musculoskeletal disorders (17.1 million) led the non-fatal burden.

More than half of the NCD deaths — 57.4% (54.2–60.4) — are attributed to the 88 risk factors the study models. High systolic blood pressure contributes the largest single share, 26.2% (21.6–30.9) of all NCD deaths in Africa, ahead of dietary risks (14.7%) and air pollution (14.2%). The authors conclude that because such a large and largely modifiable share of the burden remains, a substantial portion of it is preventable through interventions that already exist.

Where the numbers actually come from

The methods section is the part a cause-of-death reader should slow down for. The Africa estimates were assembled from an inventory of data sources that includes vital registration, verbal autopsy, surveillance systems, police reports, censuses, surveys, minimally invasive tissue sampling and sibling history, then modelled with the Cause of Death Ensemble model, DisMod-MR 2.1 and spatiotemporal Gaussian process regression.

Verbal autopsy appears in that list as a working input, not a footnote. It is the structured interview — standardised since the current WHO verbal autopsy instrument — in which a trained interviewer asks a family about the symptoms, timeline and circumstances of a death, producing a probable cause where no physician ever certified one. In much of Africa the alternative to that interview is no recorded cause at all. WHO’s assessment, carried in its civil registration fact sheet, is that nearly 40% of deaths worldwide go unregistered and that in low-income countries only about 8% of the deaths that are reported ever acquire a documented cause.

This is why a regional modelled estimate and a single household interview belong to the same story rather than to two different ones. As the baseline profile of the Bauchi surveillance site in north-east Nigeria illustrated, a model can only be as complete as the deaths that were first counted; the interview is the part of the pipeline that turns an unrecorded death into a classified one.

What the pipeline looks like on the ground

A registry in rural Punjab, reported in Cureus, shows the method doing that work for a chronic cause. Between December 2016 and November 2018, frontline health workers identified all deaths across two rural blocks of Ludhiana covering some 236,266 people, and home-based verbal autopsy interviews followed. Two independent neurologists adjudicated the stroke deaths against WHO standards and Indian Council of Medical Research VA tools, with a third expert resolving disagreements.

Stroke emerged as the leading cause of death, at 262 of 1,182 deaths — a cause-specific mortality fraction of 22.2% (95% CI 19.8–24.5). The annualised crude stroke mortality rate was 55.4 per 100,000, and 29.7 per 100,000 after age-standardisation to the WHO World Standard. Hypertension was present in 81.7% of the stroke deaths and diabetes in 34.0%; 83.6% of the deaths followed symptom onset within 30 days. The authors are candid that a VA-based diagnosis carries recall limitations, and they lean on the neurology panel’s adjudication to argue the estimate is reliable — a population-based registry in one district, not a national count.

Reading a modelled number at the right resolution

Two kinds of evidence sit side by side here, and they should not be averaged into one. The GBD analysis is a modelling exercise with wide uncertainty intervals and many collaborators; its Africa figures are estimates, not tallies. The Ludhiana study is a site-level measurement whose accuracy rests on a review panel, not on clinical certification. Different resolutions, different error structures.

A third approach sits between them. Probabilistic models such as the InterVA model, developed at Umeå University, assign a likely cause directly from the interview answers at the level of the individual death — the interview-level counterpart to a modelled regional estimate. Which tool fits depends on the question, and on whether the deaths in question were ever recorded in the first place.

What the GBD analysis makes difficult to dispute is the direction of travel: chronic causes now carry the largest single share of Africa’s mortality, and the measurement of each one still depends on systems that were, for most of the last century, built to count infectious disease and children. WHO’s own NCD fact sheet frames the broader context — that these conditions are the leading cause of death globally — but it is the African estimate that shows how quickly the gap between what people die of and what gets written down has widened.

interva.net is an independent digest on verbal autopsy, the InterVA model and cause-of-death data. It is not affiliated with, endorsed by, or speaking for the InterVA project, Umeå University or WHO.

Sources

← Back to the news digest