The Weekly Death Count and the Quarterly Cause List

New WHO guidance on rapid mortality surveillance treats cause of death as a field carrying a provenance label — medically certified, verbal autopsy (physician- or computer-coded) or family-reported — while this week's SCORE scorecard shows how few countries can fill it in.

verbal autopsyCRVSsurveillanceWHO

Syria’s public hospitals registered 12,106 deaths in 2022 through a new electronic chain built on DHIS2, an estimated one fifth of the deaths that occurred in them. In the first quarter of 2023 the quarterly figure was 6,180, an estimated 32 per cent. Both numbers describe a system learning to count faster. Neither tells you what anybody died of.

Those two figures are the worked example in a document issued by the WHO Regional Office for the Eastern Mediterranean on 12 November: Rapid mortality surveillance systems — design, operating procedures and data dissemination, a 28-page operational guide written by Chalapati Rao of the Australian National University’s National Centre for Epidemiology and Population Health with Eman Abdelkreem Hassan Aly and Henry V. Doctor of the regional office’s information systems unit. It is guidance, not research: there is no evaluation design behind it, and the two country experiences it reports — Syria, and Jordan, where a comparison of manual registration against an electronic system in six hospitals showed reporting rates 40–55 per cent higher between April and June 2023 — are described as pilots, co-run with Vital Strategies’ Data for Health grants programme, which also funded the publication.

What makes the guide worth reading closely on this beat is where it puts verbal autopsy: not at the centre of the architecture, and not at its edge, but in a specific slot on a form.

A cause field with three provenances

The document splits what a surveillance record must carry into essentials and extras. Name, date of birth and age, sex, address, national identity number where it exists, date and place of death, hospital name, date of civil registration: those are the essentials, and they are enough to produce the figure that matters in an emergency, the weekly count of deaths by age, sex and location.

The cause of death sits in the additional block, and it arrives with a question attached rather than an assumption. Was the death medically certified? If so, transcribe the certificate — every field, including durations. Was there a verbal autopsy? Then specify the source of the diagnosis: physician or computer. Or is this a cause reported by the family or some other source? Three provenances, one field. The guidance does not rank them in that table, but the structure is a quiet admission that a surveillance database will hold causes of unequal standing, and that the difference has to be recorded rather than smoothed away.

That “computer” option is where algorithmic coding lives — the InterVA model developed at Umeå University, InSilicoVA, or whatever else a ministry runs through the openVA toolchain. The guidance never names an algorithm, and it offers no accuracy figure for any of them. It simply requires the analyst downstream to know which kind of cause they are looking at, which is more than many mortality tables manage.

Two tiers, two clocks

The architecture is explicitly two-level. All-cause mortality is the primary objective and should be released weekly where the infrastructure allows, in aggregated form. Cause-specific mortality is secondary, compiled monthly or quarterly, and the guide is candid about why: medical certification is hard to obtain during the peak of a pandemic, and coding and verification take time before anything can be published. Staggered release — counts first, causes later — is how it says several countries already work, naming Brazil and Switzerland.

There is a related admission about civil registration itself. In some countries, the guide notes, deaths may be reported to the registry up to 30 days after the event, which is too slow for surveillance even where the coverage is good. The workaround is not to replace the register but to extract from it more often, which is the whole operating idea: an RMS programme is a set of activities laid over routine death recording to accelerate compilation, not a parallel system with its own forms and its own enumerators. Where the register does not reach, the guide sends the programme outward — to burial and cremation grounds, police medico-legal records, village household books, funeral services and religious authorities — and then spends a section on the consequence. The same death will be reported more than once. Its illustration is a village where the household book lists seven deaths and the health centre register six, and the true total is ten, recoverable only by matching records field by field and telephoning the field staff when a partial match is ambiguous.

Verbal autopsy enters the standard operating procedures at the community health centre, where staff are told to compile paper or electronic VA records with their causes of death, using identifying variables, and submit them with the weekly list. It is treated as one of the routine outputs of a primary reporting unit — not a research instrument bolted on afterwards. For what the instrument itself contains, the WHO 2022 verbal autopsy standard remains the reference the guidance assumes.

The scorecard underneath

Three days later, on 17 November, WHO published the second global SCORE assessment of health data systems, covering 145 countries and about 85 per cent of the world’s population. Read beside the surveillance guide, it supplies the constraint the guide is written around.

Sixty-four countries, 44 per cent of those assessed, now have what SCORE calls sustainable capacity in birth and death registration and cause-of-death reporting. No low-income country is among them. In the African Region 6 per cent of countries reach that level, against 78 per cent in Europe and 58 per cent in the Americas. Cause-of-death reporting is named the weakest domain in the whole framework: more than 70 per cent of low- and lower-middle-income countries, roughly 54 of them, do not send adequate medically certified causes of death to the WHO Mortality Database, and the global completeness of cause-of-death data is put at 48 per cent.

SCORE states the division of labour plainly. Medical certification by a trained physician is the standard; where certification is not feasible, particularly for deaths outside health facilities, verbal autopsy is the recommended method for establishing probable cause. The indicator list even carries a line item for verbal autopsy application, alongside training for certifiers and coders — a reminder that in most of the world the two routes are not competitors but adjacent columns of the same national system.

What the sequencing implies

Neither document is a study of verbal autopsy accuracy, and neither should be read as one. The surveillance guide is normative technical guidance with two descriptive pilot accounts; SCORE is a country self-assessment of system capacity, with the usual caveats about self-report and about the gap between national cause-of-death reports and what reaches the global database. What they do establish is an ordering that has become conventional: the count comes first and fast, the cause comes second and slower, and the instrument that names causes is expected to attach itself to a death that has already been counted.

For VA programmes the implication cuts both ways. Being second in the queue is not a demotion — it is the reason the method exists, since the deaths it serves are precisely those the certification route never reaches. But it does mean that a VA exercise unconnected to a death count, with no denominator and no registration behind it, is working against the grain of the guidance now being issued. The earlier entry on where verbal autopsy sits inside WHO’s own cause-of-death estimates traced the same dependency from the modelling side: the algorithm can classify, but only the registration system can say how many.


interva.net is an independent digest. It is not affiliated with the InterVA project, Umeå University, or WHO, and nothing here is medical or policy advice. Evidence levels: the rapid mortality surveillance document is normative WHO guidance with two descriptive country pilot accounts and no evaluation design; SCORE 2025 is a country-reported capacity assessment by WHO, not an independent audit; the 2022 WHO instrument is a normative standard. No accuracy estimate for verbal autopsy, InterVA or any other coding method is claimed here, and none appears in either document.

Sources

  1. WHO Regional Office for the Eastern Mediterranean — Rapid mortality surveillance systems: design, operating procedures and data dissemination (Cairo, 2026)
    iris.who.int
  2. WHO — SCORE for health data technical package: global report on health data systems and capacity, 2025
    iris.who.int
  3. WHO — 2022 verbal autopsy instrument (V2.0, multilingual)
    who.int

← Back to the news digest