WHO's New Death Count Rests on Interviews in 71 Countries

WHO's Global Health Estimates put noncommunicable diseases at 74% of deaths in 2023. The technical paper behind those figures discloses, in a table legend, that child causes for 71 countries come from models built on verbal autopsy studies.

verbal autopsycause of deathmortality dataWHO

In 2023, for the first time, communicable diseases accounted for less than half of all deaths in low-income countries. That sentence, from the World Health Organization’s latest Global Health Estimates, is the kind of milestone that gets repeated for years. It was published on 2 October and repeated near-verbatim in the November issue of the Bulletin of the World Health Organization, in a round-up of the year’s data releases.

It is also, in the strict sense, an estimate rather than a count — and where it concerns the poorest countries, it rests on interviews with bereaved families. How heavily it rests on them is described in a technical paper that most readers of the headline will never open.

What the new round reports

The estimates cover 2000 to 2023 for 183 member states and three territories. Global life expectancy reached 73.3 years in 2023, close to the 73.4 recorded in 2019; healthy life expectancy recovered more slowly, to 62.8 years, still 0.4 years below its pre-pandemic level. Noncommunicable diseases took 74% of deaths in 2023, against 58% in 2000, and eight of the ten leading causes of death were noncommunicable.

The fact sheet issued with the release puts numbers on the shift. Ischaemic heart disease killed about 9.5 million people, up 3.0 million since 2000 and 16% of all deaths; stroke and chronic obstructive pulmonary disease followed at 6.8 and 3.6 million. Lower respiratory infections reclaimed the top spot among communicable killers at 2.7 million, back to their 2019 level. Alzheimer’s disease and other dementias rose from nineteenth place in 2000 to fifth, at 2.1 million deaths, two-thirds of them women, while COVID-19 fell to fourteenth with 0.9 million. In low-income countries seven of the ten leading causes remained communicable, with lower respiratory infections first since around 2010 and diarrhoeal diseases down from first to fifth.

“Living longer is one of the great achievements of public health,” said Alain Labrique, director of WHO’s department of data, digital health, analytics and AI, in the release. “The next challenge is to ensure that those additional years are lived in good health.”

The bar a registration system has to clear

The methods paper, published in September as WHO/HSA/DDA/GHE/2026.2 and running to 58 pages, explains what had to be true for a country’s own death records to be used directly. Of the 186 country and territory rows in its annex, 66 met the inclusion criteria; for 50 of them, data ran through 2023.

The central test is a quantity WHO calls usability: completeness of death registration multiplied by the share of registered deaths that carry a meaningful cause rather than a garbage code. Data are rated high quality at an average usability of 80% or more since 2013, medium at 60%. Countries that passed every formal criterion could still be dropped for implausible cause patterns — the Philippines for some early years, Kuwait, Maldives, Suriname, Guatemala, Kazakhstan, Kyrgyzstan and Uzbekistan — or, in Bahrain’s case, because migrants were not included.

Everything else was modelled. In 119 of the annex rows, causes of death at ages five and above are labelled “GBD2023+WHO”: WHO and UN agency estimates for specific causes, set against cause fractions drawn from the Institute for Health Metrics and Evaluation’s Global Burden of Disease study, then rescaled so that the parts add up to WHO’s own all-cause totals.

Seventy-one rows and a one-line legend

For children the annex uses a different vocabulary. Its legend defines four child cause-of-death methods: death registration data; “Sample VR”, used for China and sourced from the China Disease Surveillance Points; “LMM”, models for low-mortality settings based on cause-specific mortality from vital registration; and “HMM — Models for high mortality settings based on verbal autopsy studies”.

Seventy-one rows carry the HMM label, among them India, Nigeria, Pakistan, Ethiopia, the Democratic Republic of the Congo, Indonesia, Uganda, the United Republic of Tanzania, Mozambique, Nepal, Yemen and South Africa. In other words, for the child deaths of roughly 71 countries — a large share of the world’s under-five mortality — the cause distribution is produced by models built on verbal autopsy studies rather than on medically certified records.

The phrase “verbal autopsy” appears five times in the 58 pages. The body of the paper does not spell out what the HMM label covers; that sits in the annex legend and in earlier technical papers in the same series.

WHO’s own account of what that costs

The candour is concentrated in the limitations section. Death registration, WHO writes, is “generally the best form of information available on causes of death”, but garbage codes “represent a substantial proportion of deaths in some countries, and methods for re-assigning these deaths to valid causes are generally not based on empirical data”. Assignment of an underlying cause is “quite sensitive to the order in which diagnoses are written”, variability between certifying physicians “has not been addressed to date”, and HIV and suicide are “routinely miscoded”.

Then the paragraph that matters most for anyone reading a cause-of-death table: “For many countries without functioning death registration systems, particularly in Africa, there is strong reliance on verbal autopsy studies, most of which are not nationally representative samples. Until recently there has been considerable variation in verbal autopsy instruments, and in analysis and cause assignment methods. Validation studies are challenging, and difficult to generalize to other settings.”

The paper is also explicit that WHO and IHME estimates diverge partly because IHME includes verbal autopsy studies “mapped to ICD categories using IHME-developed computer algorithms” — the family of computer-coded methods that the InterVA model, developed at Umeå University, helped establish. Different algorithms, different answers; both are documented, neither is a measurement of every death.

A revision, not a trend

One further caution belongs with the 2000-to-2023 series. This round introduces a new WHO standard population for 2026–2050, replacing the 2001 standard that was built for the years 2000–2025, and it revises both the life tables and the method used to assess completeness of registration. WHO’s instruction to users is blunt: the new series “supersedes previous GHE time series, and differences between revision series should not be interpreted as time trends”.

The trends inside a single release are internally consistent. Comparisons between two releases of the same estimates are not, and the difference between them is part of what a reader is seeing when the numbers move.

WHO’s own conclusion is that the fix is more primary data: death registration through civil registration and vital statistics systems, cause-of-death collection “through vital registration and verbal autopsy in communities”, household surveys that collect biological and clinical data, and complete facility recording with routine quality control. As we noted in an earlier entry on how child mortality estimates are built from thin evidence, the arithmetic at the top of the pyramid can only be as stable as the reporting underneath it. The 74% headline is a modelled statement about 61 million deaths; for tens of millions of them, the underlying evidence still begins with someone asking a family what happened.

Sources

  1. WHO — New WHO estimates show changing global health landscape (news release, 2 October 2026)
    who.int
  2. WHO — Global Health Estimates (data portal)
    who.int
  3. WHO — WHO methods and data sources for country-level causes of death 2000-2023, Global Health Estimates Technical Paper WHO/HSA/DDA/GHE/2026.2 (September 2026)
    cdn.who.int
  4. WHO — The top 10 causes of death (fact sheet, 2023 estimates)
    who.int
  5. Bulletin of the World Health Organization — Public health round-up, 2026;104(11), 1 November 2026
    pmc.ncbi.nlm.nih.gov
  6. WHO — 2022 WHO verbal autopsy instrument
    who.int

← Back to the news digest