> Where did COVID set routine childhood immunization back the most — and where did it bounce back?
# Vaccine equity in Sub-Saharan Africa
> Where did COVID set routine childhood immunization back the most — and where did it bounce back?
## The question
Routine childhood immunization is the single highest-leverage health intervention in Sub-Saharan Africa: every percentage point of DTP3 coverage (three doses of diphtheria–tetanus–pertussis vaccine, the standard proxy for a functioning routine programme) corresponds to thousands of averted infant deaths a year across the region. COVID-19 was the largest stress test of these systems in two decades.
This project asks, across **49 SSA countries**: where did DTP3 coverage drop the most during the disruption window, and which countries had recovered by 2024? It is **descriptive** — it quantifies and ranks the shape of disruption and recovery. It does **not** causally attribute bounce-back to specific health-system policies; that needs a different study design.
## Data
| Source | Granularity | Time coverage | Access |
|--------|-------------|---------------|--------|
| WHO/UNICEF Estimates of National Immunization Coverage (WUENIC) — DTP3, indicator `WHS4_100` | National | 2015–2024 | Public — WHO Global Health Observatory OData API |
WUENIC is the **only** data source. Access dates and raw-file provenance are pinned in `data/raw/MANIFEST.md`.
**Why national-only — a planned subnational layer was cut.** The project originally planned an admin-1 layer (DHS or the WHO/UNICEF subnational coverage dataset, mapped onto GADM polygons) to show the geography of disruption *within* countries. It was dropped: DHS microdata is login-gated and its waves mostly predate COVID, and the WHO/UNICEF subnational dataset turned out not to be freely obtainable — it is absent from the public WHO xMart API and the data portal is gated behind a JavaScript-rendered page. The only open fallback (IHME's geospatial estimates) ends before the pandemic and so cannot show the COVID *disruption* subnationally. With no usable COVID-era admin-1 …