October 4, 2026 interva.net Editorial

Zambia's Mobile Death Reporting Reaches One Death in Six

A six-month evaluation of electronic death notification across four Zambian provinces found completeness rose from 11.9% to 17.4% and reporting delays halved — while a parallel verbal autopsy programme, coded with InterVA-5, still supplies causes for deaths that never reach the register.

CRVSverbal autopsymortality datadata quality

Zambia has been trying to widen the front door of its mortality statistics — death notification — and a six-month evaluation of what happened when the official form moved onto a phone has now reported its numbers. Completeness rose, delays halved, and the register still captures about one death in six.

Published online on 28 September in Global Health Action, the study comes from Charles O. Odhiambo and colleagues at the CDC Foundation’s Bloomberg Data for Health initiative and Zambia’s Ministry of Health. It evaluates an electronic version of Form 13, the official death notification document, built in KoboToolbox and deployed at 16 sites — district burial offices and the health facilities beside them — in Lusaka, Central, Copperbelt and Southern provinces, which carry the largest volumes of death reports and the biggest paper backlogs.

A form that fits in a pocket

Many community deaths arrive at hospital mortuaries as bodies brought in for storage during funeral rites; before a body is admitted, an officer helps the family fill in the form. That document is what they need at the burial office for a burial permit and, later, a death certificate. Officers entered the details on a tablet, scanned the supporting papers, recorded cause-of-death information where it existed and synced daily; completing one form took a median of four minutes.

The platform also absorbed the backlog: paper forms for deaths from 2021 to 2023 were re-entered, while deaths from 2024 onward were captured as they happened. Over six months it held 115,573 death records covering January 2021 to June 2025.

What improved, and by how much

Completeness here is the ratio of notified deaths to expected deaths, on a denominator derived from the 2022 census crude death rate — a standard CRVS-assessment approach that the authors concede can misstate sub-national or pandemic-era mortality. On that basis the electronic system reached 17.4% (95% CI 17.3–17.5), against 11.9% (95% CI 11.8–12.0) for the manual reference period, a difference the authors report as statistically significant. Average reporting delay fell from six days to three.

The paper is careful about what the comparison can support: the paper and electronic periods were not contemporaneous, no model controlled for staffing, workload or season, and the authors call the exercise a descriptive operational comparison rather than a causal pre-post evaluation. Nearly 60% of the records came from Lusaka alone. The platform is not yet integrated with INRIS, the national registration information system, so a death can still be entered twice, in two places.

Earlier estimates put Zambia’s national death registration completeness below 15%, against a 50% target in the country’s 2021–2025 CRVS strategy; even after the improvement it stayed under 20%.

The column the study left blank

The evaluation deliberately excluded cause-of-death analysis, for the reason verbal autopsy exists: deaths in the community usually carry no medically certified cause. Where the only documentation is a police report or a family’s account at a mortuary counter, the cause on the certificate is absent or unreliable — Zambia’s earlier Lusaka work described police-recorded causes for natural deaths as very inaccurate, unreliable and unvalidated.

That is the gap the WHO verbal autopsy instrument is built for, and in Zambia it is not hypothetical: notification officers in the new system were, the study notes in passing, also conducting verbal autopsies on community deaths, and a government track has done so since 2015, when the civil registration authority sanctioned the practice.

3,411 interviews at two mortuaries

The first substantial results came from Lusaka, where a demonstration project interviewed families of bodies brought in dead at the University Teaching Hospital and Levy Mwanawasa Hospital between October 2017 and September 2018. Some 3,411 interviews were completed using the shortened SmartVA questionnaire administered on Android tablets and coded by the Tariff method; the authors estimate the process captured over 90% of deaths in the district. HIV/AIDS-related deaths accounted for 18% of adult non-facility deaths, stroke 12% and road traffic accidents 8%; among child deaths HIV/AIDS led at 19%, followed by diarrhoea and drowning; among neonatal deaths, meningitis and sepsis led at 30%, ahead of pneumonia and birth asphyxia.

Two systems then scaled up: the registration department’s across eight sites in six districts, and a Ministry of Health programme across 30 sites in 23 districts, funded through PEPFAR to monitor HIV-associated mortality. A 2025 review by the national public health institute sets out the algorithms each used — SmartVA-Analyze and InterVA-4 on the registration side, InterVA-5 on the health ministry side — and notes that all the main coding methods are available through the openVA platform. The InterVA model was developed at Umeå University and assigns causes from symptom-to-cause relationships elicited from physicians.

The same review tabulates 74,454 verbal autopsies recorded in Zambia between January 2020 and June 2024, with causes assigned by InterVA-5. Its leading categories are acute cardiac disease and “other and unspecified cardiac disease” at 11% each, HIV/AIDS-related deaths at 8.5%, stroke at 7.1%, pneumonia at 6.3% and diarrhoeal disease at 6.2%; four per cent of cases were indeterminate — symptoms too vague or contradictory for the algorithm.

The obstacles both reviews name

Neither document is sentimental about what remains. Both VA systems depend on families who bring a body to a mortuary or burial office, so deaths that reach neither — many of them rural — stay invisible. Sites were chosen purposively rather than sampled. And the VA systems are not electronically linked to CRVS or public-health surveillance, which keeps their findings out of the systems positioned to act on them. Funding is predominantly donor money; the review’s authors could not calculate an average cost per interview.

The system-level literature they lean on is older and broader: the nine-domain framework for integrating community-based verbal autopsy into CRVS set out by de Savigny and colleagues in 2017, and the lessons reported in 2021 from five countries — China, Myanmar, Papua New Guinea, the Philippines and Solomon Islands. Both treat verbal autopsy less as a stopgap than as a component to be designed into registration systems rather than bolted on afterwards.

What the Zambian evaluations add is arithmetic. Digitising notification halved the reporting delay and moved completeness by roughly the same proportion — and more than four in five deaths still never enter the register. The cause-of-death data that does emerge from community deaths comes largely from interviews with bereaved families, run through probabilistic models, in places where no physician was ever in the room.

Sources

← Back to the news digest