October 12, 2026 interva.net Editorial

Three of Fourteen: How Death Notification Found Uganda's Sudan Virus Cases

In Uganda's 2025 Sudan virus disease outbreak, three of the fourteen cases entered the surveillance system through a death alert — including the index case, confirmed after death. An operational review published in September describes how mortality surveillance worked; a 2025 assessment of Uganda's routine death reporting shows how much of the country's mortality still runs past the desk.

surveillancemortality dataverbal autopsyCRVS

Uganda’s sixth outbreak of Sudan virus disease was declared on 30 January 2025 and declared over on 26 April — eighty-seven days from the first confirmed case to the end of transmission. The virus behaved the way filoviruses now tend to behave in the region’s cities: it surfaced in Kampala’s metropolitan area, where people move between districts, seek care from several providers, and where a large share of deaths happen at home in informal settlements rather than in a hospital bed with a chart attached.

Fourteen people met the case definition — twelve confirmed in the laboratory, two classified as probable. By 5 March, WHO’s outbreak notice counted the same fourteen cases and four deaths. What that notice could not show is the route each case took into the surveillance system. Three of the fourteen are documented as having arrived through a death.

The four steps a death has to travel

The account comes from an operational review published on 21 September in the Journal of Epidemiology and Global Health by Rogers Moses and colleagues from WHO’s Uganda country office, the Ministry of Health’s Department of Integrated Epidemiology, Surveillance and Public Health Emergencies, and Makerere University’s School of Public Health. Its evidence base is documentary: situation reports, incident-management meeting minutes, technical briefs, alert and call-centre logs and mortality surveillance line lists, all produced between the declaration and the end of the outbreak. The authors are explicit that it is not an impact evaluation, and that it cannot separate the effect of mortality surveillance from the case-finding, contact tracing and safe-burial work running beside it.

Their working definition of mortality surveillance has four parts, and the sequence matters. First, systematic detection, notification and verification of deaths. Second, investigation of those deaths, including verbal autopsy and epidemiological linkage. Third, post-mortem specimen collection and laboratory diagnosis. Fourth, the public health action a death triggers — contact tracing, quarantine, safe and dignified burial.

In practice the first step ran on people who were already in the community. Families and neighbours reported unexplained, non-traumatic deaths — fever, bleeding, sudden deterioration — to community health workers or to toll-free hotlines; the health workers counselled the family and relayed the alert to a district Ebola rapid response team. In Kampala’s informal settlements, where many deaths occur at home, that channel was the only one likely to reach them. Inside facilities, clinicians, pathologists and mortuary attendants telephoned alerts directly. An enhanced alert management system was activated in the Kampala metropolitan area, Jinja and Mbale, overseen by a national alert desk at the ministry and wired into the national call centre, with digitised alert forms and call logs feeding a single repository. Across the response, the review records 3,757 notifications or signals and 2,700 alerts.

Then came the interview. Rapid response teams conducted on-site investigations that included what the paper calls structured verbal autopsies with family members or health workers, documenting symptoms and exposures — contact with a confirmed case, attendance at a funeral — which fed the classification of a death as suspect, probable or unrelated to Sudan virus. This is the step that keeps the rest of the machinery honest, and it is recognisably the same instrument logic that the WHO’s 2022 verbal autopsy instrument formalises for routine use, where a structured interview is the input to algorithms such as the InterVA model and its data-driven alternatives and emerges as a coded cause of death. In an outbreak, the same interview is used for triage rather than coding — and the review notes that it was the verbal autopsy step which fed directly into contact tracing and case investigation.

What the death alerts produced

Three of the fourteen cases had a death investigation as their documented entry point into surveillance — the index case, confirmed after death, and cases found when a death alert triggered investigation, specimen collection and laboratory testing. One of them was a secondary case in a healthcare worker epidemiologically linked to the index case. The authors chose the conservative reading: where the records did not make clear whether a case was detected solely through the mortality pathway or during a broader investigation that a death had started, they counted it by the documented entry point. Their figure of 21.4 per cent is therefore a floor.

The death investigations did other work too. They contributed 112 of the 534 contacts listed during the response — about a fifth — and informed 167 safe and dignified burials. For those deaths, the review puts the typical interval from post-mortem notification to laboratory confirmation at 24 to 48 hours, and notes that earlier filovirus outbreaks in the region reported longer delays from symptom onset or death to detection, while cautioning that those figures come from different settings and are not a controlled comparison.

The friction points are as instructive as the successes. Deaths went unreported from informal settlements and private clinics, where care is often sought from unregulated providers or given at home — the same blind spot that any community-based death-notification system meets. Sample transport for testing was slow in places, and swabbing was delayed beyond 48 hours for a proportion of notifications, sometimes because families hesitated over stigma. And there was no mortality surveillance guideline in place during the response at all. Districts and partners applied inconsistent case definitions and data practices, which the authors say hampered aggregation and comparison and reduced the sensitivity of the system. Uganda’s national mortality surveillance guidelines were finalised only in May 2025 — after the outbreak had been declared over.

Two-thirds of Uganda’s deaths never reach a desk

How much room there is to improve is the subject of a separate Ugandan paper. In PLOS Global Public Health in July 2025, Miriam Zalwango and colleagues from the Ministry of Health’s integrated epidemiology department reported a rapid assessment of mortality surveillance carried out between November 2023 and June 2024, built on eight workshops and meetings, thematic analysis of stakeholder discussions and an analysis of district health information system data.

Its headline number is geography, not method. Some 66.9 per cent of deaths in Uganda occur in communities, and the data generated there reaches the national health statistics database in DHIS2 only suboptimally. The remaining third of deaths occur in health facilities, and that stream has problems of its own: national referral and specialised institutes not feeding their data into the national database, late reporting, and a lack of coordination and standardisation between the partners doing the reporting. The authors’ prescriptions are institutional — clearer governance of death-reporting, defined roles and responsibilities, awareness raising, technical capacity, resources, and alignment of partners behind one system — rather than technical.

Set the two papers side by side and the contrast is sharp. In an outbreak, with an incident-management structure, dedicated hotlines, digitised alerts, trained rapid response teams and community engagement that oriented more than 1,500 community health workers and 363 community leaders, deaths became legible within a day or two of occurring, and at least three cases exist in the surveillance record because of it. In routine years, roughly two-thirds of deaths occur in places the register does not reliably reach. A 2025 Bulletin of the World Health Organization paper by Chalapati Rao and colleagues argues for treating this routine mortality surveillance as core pandemic preparedness rather than an emergency improvisation; WHO’s own CRVS fact sheet puts the global gap at close to 40 per cent of deaths unregistered, and reports that in low-income countries only about 8 per cent of reported deaths have a documented cause.

The Zambia question is the same question: digitising death notification at burial offices and mortuaries raised registration completeness but did not, on its own, produce causes of death. Uganda’s 2025 response did something narrower and faster — it made deaths visible in time to act on them, without attempting to code them. Both are worth reading as what they are: a descriptive review of one outbreak’s death-notification chain, and a rapid institutional assessment, neither of which establishes that any particular system, run any particular way, would have stopped transmission earlier.

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