BAY-States Cholera: the DRC Trigger Applied to Borno, Adamawa and Yobe
Suspected cholera surveillance in Borno, Adamawa and Yobe (2018–2023), and what the DRC cholera framework’s 2026 trigger would have done with it
Summary
This is an analysis, not a framework in development. It asks whether the Borno–Adamawa–Yobe (BAY) suspected-cholera line-list could support an anticipatory action trigger built like the one in the DRC cholera framework. We applied the method of that framework’s 2026 revision (endorsed June 2026) to the line-list, as we did for Haiti in October 2026, and compared the results with CERF’s cholera allocations to Nigeria.
This version replaces the earlier one (July 2026), which tested a design drafted for BAY in 2024: a threshold per head of population and no spatial constraint. That design gave 27 LGA-year activations. The comparison is in The DRC trigger applied to BAY.
Closer than Haiti, but not yet. The line-list covers six years (32,826 suspected cases, 2018–2023, with 2020 missing) and three large outbreak years. Applied as it stands, the DRC rule would have activated 2 times: Gwoza (Borno), week ending 5 September 2021, and Jere (Borno), week ending 2 October 2022. It activates before the 2021 CERF cholera allocation, by three weeks on onset dates. It does not activate in the 2018 wave that CERF funded; of the thresholds tested, only the DRC’s lowest (63) does, in Yobe. And it activates in 2022, the largest outbreak, which drew no CERF allocation. The line-list ends in December 2023, has no laboratory results, and does not show when cases were reported. Those three gaps, not the method, are what stand between this and a trigger.
What we found
The record. 32,826 suspected cases and 836 deaths in 55 LGAs, 2018–2023. Adamawa had the most cases in 2018, and Borno in both 2021 and 2022. 2020 is a reporting gap (3 cases), not a quiet year, and is left out of every baseline (Data quality).
The threshold. The DRC rule sets the threshold each January at the 98.8th percentile of weekly cases in units with at least one case, over the last five years. For BAY in January 2024 (2018–2023 without 2020) that gives 251 suspected cases in an LGA in a week, above the DRC’s 2026 endemic threshold (174). Taking each year’s threshold from earlier years only gives 113 to 269 (The DRC trigger applied to BAY).
Would-have-activated history. With 251, 4 LGA outbreaks reach three alert weeks, all in Borno, and the L3 constraint keeps 2: a return period of 3 years over the five usable years. A lower threshold (the expanding baseline, or the DRC’s 174) adds Mubi North (Adamawa) in May 2018. The DRC’s non-endemic threshold (63) gives 5 (Activation frequency).
With three states, the L3 constraint mutes nearly everything. Borno borders both Adamawa and Yobe, so an activation in Borno mutes all of BAY for six months. A same-state constraint (the DRC team’s L2.5 alternative) would be the closer equivalent of the DRC’s 26 provinces. It changes nothing under the reference threshold, and adds two activations under the 63 threshold.
The rule activates at the peak of each LGA’s outbreak, not before it. BAY outbreaks rise fast, and three alert weeks are needed: most activations come in the LGA’s peak week or the week after, when 45–62% of its cases have happened. At state level it can still be early: Gwoza came five weeks before Borno’s 2021 peak.
CERF. Two CERF cholera allocations fall inside the record, both responses to national outbreaks. The rule activates before the September 2021 one under every baseline (22 to 36 days, on onset dates). For October 2018, only the 63 threshold activates in the August–October wave the allocation responded to (Damaturu, Yobe, 52 days before). The expanding baseline and the 174 threshold activate only on a separate Adamawa outbreak in May, and under L3 that mutes Adamawa and Borno for the autumn wave. 2022, the largest outbreak (12,469 suspected cases in Borno), drew no CERF cholera allocation. The May 2026 allocation for Borno is after the record ends (Comparison with CERF allocations).
What cannot be tested. The 2026 framework requires a laboratory-confirmed case in the 30 days before activation; the line-list has no laboratory results. Cases are dated by onset (95% are seen at a health facility within 4 days), but not by the date they reached the surveillance system. A separate IDSR file shows BAY LGAs’ weekly reports arrived on time in 86–98% of weeks in 2023 and early 2024.
Is the data enough?
| DRC (2026 revision) | Haiti (Oct 2026 look) | BAY (this data) | |
|---|---|---|---|
| Weekly data | 20 years (IDSR 2006–2025) | 4 years (Sep 2022 – Sep 2026) | 6 years (2018–2023), 2020 missing; ends Dec 2023 |
| Units | ~515 health zones, 26 provinces | 140 communes, 10 departments | 55 LGAs, 3 states |
| Outbreak periods to learn from | Many | One national wave | Three (2018, 2021, 2022) |
| CERF cholera allocations to check against | 8 evaluable | 1, at re-emergence | 2, both national responses; a third (2026) after the data end |
| Threshold stability | Rolling 5-year window, coefficient of variation ~19% | 115 or 35, depending on the window | 113 to 269 by year (expanding) |
| Missing reports | 2020–21 gap found and excluded | Fully zero-filled; no way to find gaps | 2020 gap found and excluded; no completeness flag in the line-list |
| Laboratory confirmation | Condition in the framework | Confirmed cases in the data | None in the line-list |
For an escalation trigger, BAY’s record is better than Haiti’s: more outbreaks, two CERF allocations, and a threshold that does not swing by a factor of three with the window. It is still short, it stops in 2023, and the two conditions the DRC framework relies on in operation — weekly reporting with a known delay, and laboratory confirmation — are not in it.
What would help
From NCDC, the state ministries of health and WHO:
- Weekly LGA counts for 2024–2026, ideally from the IDSR system an operational trigger would use. This would add the May 2026 CERF allocation as a third event to check against, and allow the January threshold to be computed on current data.
- Laboratory results by LGA and week (RDT and culture), so the framework’s confirmation condition can be tested.
- The date each case was reported to the state or NCDC, to measure the delay between onset and when a trigger could see it.
- Any list of cholera hotspot LGAs NCDC or the states already use. A fixed list, like the DRC’s endemic provinces, would allow the threshold to be split.
- What happened to the 2020 data. Outbreaks were reported in all three states that year.
And from us:
- What should a BAY trigger catch? Outbreaks of the 2022 size, which drew no CERF money, or only those that did? The CERF record cannot answer that.
- How much spatial muting? With three states, L3 mutes BAY for six months after any Borno activation. L2.5 is the obvious alternative.
- Which threshold? The DRC procedure gives 251; the DRC’s own 2026 values (174, 63) would activate more often and earlier.
Data sources
| Source | What | Coverage | Notes |
|---|---|---|---|
CHOLERA_BAY 2018_2023.xlsx |
Case line-list (one row per suspected case) | Borno, Adamawa, Yobe · 2018–2023 | Onset date, date seen at a health facility, LGA, ward, outcome, age, sex, OCV. No laboratory results. Blob pa-aa-nga-cholera/raw/healthsector/. |
nga-adm2-populations-2023.csv |
2023 LGA populations | 55 BAY LGAs | Defines the LGAs in the panel. |
COD-AB nga_adm2 |
LGA boundaries | Nigeria | Maps and state adjacency. |
| IDSR completeness and timeliness | Whether each LGA’s weekly IDSR report arrived on time | BAY · 2023 and 2024 to week 15 | Blob nga-aa-cholera/private/raw/cholera/. |
| CERF OneGMS | Cholera allocations to Nigeria | 2018, 2021, 2026 | Public feed, read 4 November 2026. |
| DRC trigger code | The 2026 revision’s L3 rule | — | pa-aa-cod-infectious-disease via the Haiti port (Method notes). |
Chapters
- 1 · Data overview: what the line-list holds; coverage by year, state and LGA; outcomes.
- 2 · Data quality: reporting over time and the 2020 gap.
- 3 · Epidemiology: epidemic curves, seasonality, where the cases are, who they are.
- 4 · The DRC trigger applied to BAY: the rule, the thresholds, weekly levels, onsets and constraints; comparison with the earlier design.
- 5 · Activation frequency: activations by baseline and year, return periods, sensitivity, timing within outbreaks.
- 6 · Comparison with CERF allocations: 2018, 2021, 2022 and 2026; reporting delays.
- Method notes: the port of the DRC code, choices for BAY, what was not done.
Glossary
- BAY: Borno, Adamawa and Yobe states, north-east Nigeria.
- LGA: Local Government Area (admin level 2), the unit here.
- Suspected case: a case meeting the clinical cholera definition; not necessarily confirmed by a laboratory.
- Alert week: a week in which an LGA’s cases are above the percentile threshold, or more than four times the previous week’s.
- Onset (L1): the third consecutive alert week in an LGA.
- L2: onsets left after muting an LGA for six months after its own activation.
- L3: onsets left after also muting, for six months, the state of each activation and its neighbouring states. The DRC framework’s configuration.
- L2.5: as L3, but muting the state only.
- Activation: an onset kept at L3. “Would have activated”: in the backtest, not in reality.
- Return period: (years + 1) / years with an activation (Weibull plotting position).
- CERF: the UN Central Emergency Response Fund; its Rapid Response window funded all three cholera allocations to Nigeria.