Four Boxes for Sixty-One Thousand Deaths

Inside Sudan the best available mortality estimate rests on death lists and a four-way cause question. Comparing that with what a full verbal autopsy produces in Somalia and Chad shows where the record stops.

verbal autopsymortality datadata qualityglobal health

The questionnaire behind the most detailed mortality estimate yet published from inside Sudan’s war asked the respondent to pick one cause of death from four options: accidental injury, intentional injury, starvation or disease, or other. Not a symptom list. Not the long interview schedule of the WHO verbal autopsy instrument. Four boxes and a name.

That is the cause-of-death vocabulary underneath a paper in The Lancet Global Health, which estimated 61,202 deaths (95% CI 22,286–209,151) in Khartoum State between April 2023 and June 2024, of which 26,024 (95% CI 12,571–58,704) were attributed to intentional injuries. Maysoon Dahab and colleagues at the London School of Hygiene & Tropical Medicine were not cutting corners. They were working in a country whose civil registration had already been too weak to count deaths reliably before the fighting began in April 2023, and whose contested areas then lost what remained.

What a list of names can carry

The study’s method is capture–recapture: estimate a hidden total from the overlap between incomplete samples. Three lists of dead people were assembled remotely: a public survey pushed through social platforms, a private survey spread by chain referral, and obituaries scraped from roughly 150 social media pages. Some 417 respondents completed 465 questionnaires; the lists together held 6,715 records, which after de-duplication and cross-list matching resolved into 4,884 unique decedents, 3,846 of them with enough detail for descriptive analysis.

Multiple systems estimation then models the overlap between the lists — how many people caught by one list were also caught by another — and infers how many deaths appeared on none of them. The answer in Khartoum was 95%. Almost every death the study estimates never reached a single list.

The point estimate carries a confidence interval that spans roughly a factor of nine, and the authors are explicit about why: when they varied the thresholds used to decide whether two records from different lists described the same person, estimates ranged from about 21,000 to 211,000 all-cause deaths. Record linkage in a country where a handful of given names recur across millions of people is not a bookkeeping detail. It is the analysis.

The ratio nobody can check

On 2 November, BMJ Global Health published a commentary by Debarati Guha-Sapir and Thomas Jideofor Ogbu arguing that counting the dead in Sudan is a moral act as much as a statistical one — that credible figures “protect suffering populations from statistical invisibility” and lay a foundation for accountability. It is a commentary, not new data, and its numbers are drawn from humanitarian and media reporting: more than 150,000 killed since April 2023, nearly 14 million displaced, up to 80% of health facilities in conflict-affected areas out of service.

Its most consequential move is arithmetic. Citing a Small Arms Survey estimate of four indirect deaths for every direct one, the authors apply that ratio to the 18,916 direct deaths recorded by ACLED and conclude that indirect deaths “may exceed 75,000” in Darfur alone. The ratio is imported from other wars. It cannot be tested here, because testing it requires knowing what people actually died of — hunger, cholera, measles, a bombed clinic, a bullet — and no system in Sudan is producing that.

The Lancet estimate already complicates the denominator. Its 26,024 intentional-injury deaths for Khartoum State alone exceed the 20,178 conflict fatalities ACLED recorded nationwide for the same fourteen months. If the direct count is that incomplete, a ratio applied to it inherits the incompleteness.

Where the instrument has been run

Verbal autopsy exists precisely for the setting Sudan is now in. A structured interview with a bereaved family member collects symptoms, timing and care-seeking; a physician panel or a probabilistic model then assigns a probable cause, either through physician review or an algorithm such as the InterVA model developed at Umeå University. The WHO’s current instrument runs to well over a hundred questions across age-specific modules — the machinery behind what a verbal autopsy actually measures, and the reason algorithmic assignment can say “neonatal pneumonia” rather than “other”.

Two nearby studies show what that machinery delivers when it can be deployed. In Somalia’s Banadir region, researchers combined community-based surveillance with verbal autopsy in 57 camps for internally displaced people: 20,323 residents enumerated, 722 deaths recorded, and the WHO questionnaire administered by two medical doctors, with causes assigned by InSilicoVA through the openVA package. Crude death rates in Daynile and Kahda stood at 3.15 and 1.26 per 10,000 person-days in the October 2022–February 2023 window, then fell to 0.46 and 0.34 by March 2023 as outreach teams arrived. Severe malnutrition, respiratory infections and diarrhoeal disease led the cause list; among under-fives, malnutrition, measles and neonatal pneumonia.

And a result worth pausing on: the study recorded no violence-related deaths at all, despite running inside an active conflict. The authors attribute this partly to security inside the camps, and partly to the possibility that fear or stigma suppressed reporting. A questionnaire administered by doctors still returns what families are willing to say.

What ground access changes

Across the border, three retrospective household surveys in eastern Chad in August and September 2023 reached refugees and returnees from West Darfur. These collected self-reported causes, not verbal autopsies, and they had something the Sudan study lacked: interviewers in the camps. Violence, specifically shootings, was the leading reported cause in all three sites. In Ourang camp, where more than 90% of households came from El Geneina, the crude death rate during the crisis phase was 2.25 per 10,000 person-days against 0.11 before it — a twentyfold rise — and more than one in ten men aged 30 and over died of violent causes. Across the three camps, 2.4% of household members from El Geneina died violently, nearly five times the share among households from other places of origin.

These are the same problem of who gets counted that our October piece on Burkina Faso’s phone surveys described from a different angle: in insecure regions the sampling frame, not the questionnaire, decides what a mortality estimate can say.

What the record is for

The ICC’s conviction of Ali Muhammad Ali Abd-Al-Rahman in October 2025, on 27 counts of war crimes and crimes against humanity in Darfur two decades earlier, is the backdrop the commentary invokes: justice arriving years after the dead were counted, if they were counted at all. Whether today’s toll is documented well enough to support a future reckoning depends on instruments that require a household list, physical or telephone access, trained interviewers, and time — none of which a besieged capital supplies.

The Lancet authors say as much in their limitations. Incomplete information on the circumstances of individual deaths, they write, restricted their ability to characterise those deaths and to infer intentionality. That sentence is the whole distance between a count and a cause. Sudan can now say, within a wide interval, how many people died in its capital. Four boxes is what it can say about why.

Sources

  1. Guha-Sapir D, Ogbu TJ. No peace without accountability: mortality and public health in Sudan's war. BMJ Glob Health 2026;11(11):e022673 (commentary, no new data)
    pmc.ncbi.nlm.nih.gov
  2. Dahab M, AbuKoura R, Checchi F, et al. War-time mortality in Sudan: a multiple systems estimation analysis. Lancet Glob Health 2025;13(9):e1583-e1590 (open-access copy, LSHTM Research Online)
    researchonline.lshtm.ac.uk
  3. Simons E, Ouedraogo P, Garnier C, et al. High mortality and violence among refugees and returnees from West Darfur, Sudan arriving in Chad, 2023: results from three retrospective mortality surveys. BMJ Glob Health 2025;10(6):e016668
    pmc.ncbi.nlm.nih.gov
  4. Community-based mortality surveillance among internally displaced vulnerable populations in Banadir region, Somalia, 2022-2023. Front Public Health 2025;13:1582558 (WHO verbal autopsy with community-based surveillance)
    pmc.ncbi.nlm.nih.gov
  5. WHO verbal autopsy instrument (2022 revision), World Health Organization
    who.int

← Back to the news digest