October 8, 2026 interva.net Editorial

The Count Before the Cause: A New Surveillance Ledger in North-East Nigeria

A baseline profile of the Bauchi health and demographic surveillance site — 151,672 people, 23,476 households — is a reminder that verbal autopsy can only assign a cause to a death somebody first recorded.

verbal autopsyCRVSsurveillancechild mortality

Verbal autopsy answers a narrow and specific question: given that this person died, what most likely killed them? Everything about the method — the questionnaire, the algorithms, the long negotiations over who reviews a contested case — sits downstream of a prior fact. Someone had to notice the death, write down who it was, and place it on a calendar. In much of the world that earlier step is where the record breaks.

WHO’s current assessment puts the scale of the leak plainly: nearly 40% of deaths worldwide go unregistered, and in low-income countries only about 8% of the deaths that are reported ever acquire a documented cause. Across the African Region, the share of deaths entering a civil registration system is roughly one in ten. No cause-of-death tool, however well validated, can classify a death that never entered a ledger.

That is the gap a paper in Global Health Action sets out to narrow, from the north-east of Nigeria rather than from a global estimate.

A ledger for eight wards

The Bauchi Health and Demographic Surveillance System was established in October 2023 with support from the CHAMPS network — the Child Health and Mortality Prevention Surveillance programme — and is run from Abubakar Tafawa Balewa University. Its baseline enumeration covered 257 enumeration areas across eight contiguous administrative wards in the Bauchi and Ganjuwa local government areas, split between an urban catchment inside the state capital and a rural catchment around Kafin Madaki, some 30 kilometres away.

The first census registered 151,672 people living in 23,476 households across a surveillance area of roughly 591 square kilometres. With it the site acquired the two things population work cannot proceed without: a denominator and a follow-up schedule. Two rounds of six-monthly data collection now track births, deaths, migration and household change — the machinery that turns scattered interviews into rates.

The authors are explicit about what the platform is for. Alongside fertility and migration, they describe it as a base for pregnancy and mortality surveillance and for verbal and social autopsy, and as a setting for future studies nested inside it. That is the standard arrangement: a surveillance site supplies the deaths and the population at risk, and the cause-of-death work happens on top. The research projects built on such platforms are, in effect, the reason for building them.

What the baseline found

The baseline profile is mostly a portrait of who lives in the catchment, and it reads as a young and thinly provisioned population. Just over half — 53.3% — are under 15. Children under five number 25,059, or 16.5% of the total, while people aged 65 and over make up 2.4%. Women of reproductive age total 39,175. The dependency ratio stands at 104, meaning there are slightly more dependents than working-age adults, and the mean household size is seven.

Economic indicators are stark, though they describe the enumerated population rather than the whole state. Among residents aged 12 and older, most have no formal education; unemployment sits at 48.97%, with only 7.12% in full employment and 40.63% self-employed.

Household conditions carry the clearest health signal. Boreholes supply 31.13% of drinking water, protected wells 26.8% and unprotected wells 21.07%. Sanitation rests on pit latrines, 64.9% of them inside the compound. For cooking, 78% of the population burns wood, grass, straw or brush, and a further 13.4% uses charcoal — a near-universal reliance on solid fuel that the authors connect to indoor air pollution and respiratory illness.

The surrounding mortality arithmetic gives those numbers their weight. Citing the 2022 Multiple Indicator Cluster Survey and the 2023 Nigeria Demographic and Health Survey, the paper notes an under-five mortality rate of 127 deaths per 1,000 live births in the North-East, above the national figure of 102. For Bauchi State itself it lists 147 under-five deaths, 69 infant deaths and 38 neonatal deaths per 1,000 live births, and a maternal mortality ratio of 1,549 per 100,000 live births. Only 21% of deliveries take place in a health facility.

What it took to build

The operational section is the part worth reading twice, because it is where most surveillance ambitions quietly fail. The single largest obstacle was cartography: the teams worked from enumeration area maps last updated in 2005, and by 2023 settlements had outgrown them. The result was omitted communities, overlapping boundaries and duplicate households in the listing. A supplementary enumeration of 44 further enumeration areas ran between May and July 2025 to reach locations the original frame had missed.

Two other problems recur. Early community mistrust and limited understanding of what a surveillance site is for depressed participation until community liaison officers were recruited to mediate rejections; the paper treats sustained local engagement as the foundational fix rather than an add-on. And the data system itself was unstable in the first months — crashes, locked-down device options and unreliable GPS capture — to the point that the data manager and programmer were sent for training in Kisumu, Kenya before collection stabilised.

Data quality was handled with layered checks: SQL queries that automatically flagged errors in uploaded records, supervisor back-checks on sampled households, and unannounced spot checks on field staff. Seasonal farming migration required follow-up visits to households that were absent during enumeration.

What a platform like this can and cannot say

The paper’s honesty about its own boundaries matters as much as its numbers. Because clusters were drawn by proximity to health facilities, some communities in the selected wards were left out, and the authors state that the surveillance area may not represent the wards as a whole. The study is a descriptive baseline, not a validation exercise: no causes of death are reported here, and verbal autopsy appears as a planned capability rather than a result. No accuracy figure can be taken from it — not for physician review, and not for a probabilistic model such as the InterVA model developed at Umeå University.

It is also only one arm of CHAMPS as the network describes itself. The programme’s better-known output pairs interview-based classification with minimally invasive tissue sampling to name pathogens in individual child deaths — the investigation work reported this year across seven countries, and far more precise than what a demographic platform alone can produce. The two are complements, not rivals: the tissue-sampling route establishes what killed a particular child, while a site like Bauchi establishes how many children die, where, and against what backdrop of deprivation.

The distinction is easy to lose in summaries that treat “cause of death data” as a single object. A surveillance site and a cause-assignment algorithm answer questions at different resolutions, and the field needs both. What the Bauchi baseline quietly supplies is the precondition for the more publicised half: a list of deaths accurate enough that a verbal autopsy interview has somebody to be about.

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