Before the Interview: Uganda Moves Death Notification Into the Village Health Worker's Phone
A five-day, 26-person hackathon in Uganda welded community death notification into the app village health teams already use — a fix aimed at the step before verbal autopsy, where deaths currently vanish between paper forms, police ledgers and the civil registry.
Uganda’s village health teams have been reporting deaths for years, on paper, quarterly, into forms that travel from a village to a health facility and then into a district database. Sometimes the reports arrive twice. Often they do not arrive at all. Between January and August 2021, only 39% of the deaths reported through national surveillance originated in the community, in a country where roughly six in ten deaths happen outside a health facility.
That arithmetic is the subject of a paper published in April in Scientific Reports by a team from Uganda’s Ministry of Health surveillance department, the Uganda National Institute of Public Health and the national identification authority. It describes a deliberately modest intervention: rather than build a new system, they spent three workshops in 2024 redesigning the moment a death is first reported, so that it happens inside the mobile app that village health teams already carry.
The step that comes before the interview
Verbal autopsy gets the attention in this field — the questionnaires, the algorithms, the arguments about how closely a computer-coded cause matches a physician’s reading or a tissue sample. All of it sits downstream of something duller and more fragile: a household death has to be noticed, recorded, and passed into a system that will eventually send an interviewer to the family.
Where that notification step fails, no amount of methodological sophistication recovers the death. The WHO estimates that nearly 40% of the world’s deaths are never registered, and that only 8% of reported deaths in low-income countries carry a documented cause; its latest capacity assessment put registered deaths at about 10% in the African Region. The WHO’s 2022 verbal autopsy instrument, the field’s current normative tool, is built to be administered to a family that has already been identified as bereaved.
The Ugandan team was explicit that their work is not verbal autopsy. The paper says it does not address “comprehensive CRVS reform, verbal autopsy implementation, or medical certification of cause of death.” That framing is worth taking at face value — and it is exactly why the study is interesting.
Five days, twenty-six people, and a wall of parallel forms
The design was a participatory qualitative study wrapped around a hackathon: two three-day sessions and one five-day session between May and August 2024, two in Kampala’s Makindye Division for the urban case and one in Masaka District for the rural one. Twenty-six people took part. Twelve were village health team members, split into two focus groups of six; the rest came through fourteen in-depth interviews with ministry surveillance officials, three staff from the National Identification and Registration Authority, three facility records officers, two police officers who handle death investigation, and two developers of the eCHIS platform.
What they mapped first was duplication. Village health teams file deaths on HMIS 007 for maternal deaths, HMIS 008 for child deaths, and the community health summary form 097b, submitting quarterly to a health facility that aggregates them into DHIS2. A team member in Masaka described the loop as one-directional: the forms go up, and nothing comes back. “We get no feedback and don’t know what they do with this data,” the participant said. An earlier experiment during COVID-19, an SMS short code for suspected pandemic deaths, was never wired into DHIS2 and stopped when its funding ended — a pattern one Makindye volunteer summarised as reporting into a research project that then disappeared.
The civil registration side runs on its own documentation logic. Registering a death with the national authority requires Form 12 and entry into the Mobile Vital Registration System, plus family meeting minutes naming the person authorised to process the certificate, photocopies of identity documents and signatures; deaths before 1 January 1990 need a recommendation letter endorsed by the local council chairperson and the district internal security officer. For a community death, the declarant has to appear in person at the parish or subcounty office. Records officers at a regional referral hospital added a third variant of the same problem: unclear rules for deaths in emergency units and outpatient departments, and for referred patients who died before registration, leaving staff unsure whether a death belonged in an inpatient or outpatient count.
Police registers and mortuary books hold yet another copy of the same event, shared only on request. Nothing in the study suggests anyone was reconciling these streams.
What the new form refuses to ask
The workflow that came out of the workshops is intentionally shallow. It captures age, sex, date of death, place of death, and manner of death — disease-related, accident, suicide, homicide or other external cause — plus screening questions for epidemic-prone conditions and acute fatal events such as road traffic injuries. It does not ask a village health worker to name a medical cause, on the explicit reasoning that they are not clinically trained to do so. Cause assignment stays with facilities and the certification system; what the community layer contributes is the existence of a death and its rough character.
Technically, the configured form synchronises with DHIS2, where dashboards break notifications down by age, sex, district and manner of death. At the time of configuration, eCHIS covered about 11,099 village health teams and 160 community health extension workers across 17 districts. The authors report that early feedback from participating districts pointed to more consistent reporting and less duplication between channels — while stating plainly that a formal quantitative evaluation of completeness, accuracy or timeliness was outside the study’s scope.
That caveat matters more than the pilot signals. This is a qualitative, purposive study of 26 people in two locations, analysed through reflexive thematic analysis with coding done in spreadsheets, and “consensus” defined as the absence of substantive objection after moderated discussion. Several authors sit inside the ministry division and the eCHIS implementation team, which they acknowledge and address by having two analysts code independently. And the interoperability the paper is really about — reconciliation between the ministry’s platforms and the registration authority’s system — was not implemented or tested, only architected.
Why the cause-of-death world should read it
Two things travel beyond Uganda. The first is the reminder that verbal autopsy coverage is capped by notification coverage. A country can run a well-validated instrument, a documented coding pipeline and a clean cause list, and still produce a cause-of-death distribution for a biased subset of its dead. Debates about whether the InterVA model or a physician panel reads a case better are, in that setting, second-order.
The second is the shape of the fix. Zambia’s digitisation of its death-notification form, reported in this digest earlier, went after the same front end by putting a paper form into a mobile data tool. Uganda’s version keeps the form deliberately thin and bets on the phone already in a volunteer’s hand, which is cheaper but also puts weight on supervision and feedback — the paper’s own discussion concedes that digital tools without feedback loops and district-level data use are unlikely to change much.
Notification is where a death becomes countable. Until it happens, the interview, the algorithm and the certificate are all waiting for a family nobody has told them about.
Sources
- Alunyo JP, Bulamu M, Akello SR, et al. Enhancing community death notification in Uganda using a hackathon based approach to integrate mortality surveillance into the electronic Community Health Information System (eCHIS). Scientific Reports 2026;16:17227. pmc.ncbi.nlm.nih.gov
- World Health Organization. Civil registration and vital statistics (fact sheet). who.int
- World Health Organization. 2022 WHO verbal autopsy instrument. who.int