Africa/Methodology
Sources & method

How Veligraph is built

255,963 facility records across 53 countries, rebuilt from official sources into one schema. This page is the method, plainly stated.

Sources

Records come from four classes of source, in order of preference:

Every record stores its source name, vintage and licence. We name sources on every page and never scrape private or paywalled databases.

Pipeline

  1. Normalize. Each source is mapped into one schema of 38 attributes: identity, place, level of care, ownership, capacity, staffing, services, licensing, coordinates.
  2. Classify. Facility type and ownership are classified with multilingual rules (English, French, Portuguese) when the source does not state them.
  3. Deduplicate. Exact duplicates are removed, then near-duplicates are collapsed: same name in the same state and district is treated as one facility, keeping the most complete record; same-name facilities in different districts stay separate.
  4. Flag, never edit. Implausible figures (for example a bed count far above the facility's registered level, or more doctors than beds) are flagged for verification and shown flagged. We do not silently change source numbers.

How to read our counts

Counts on this site are facilities tracked by Veligraph, not national totals. Coverage varies by country with the quality of the available sources, and the private sector is undercounted in several countries. Where we know coverage is partial, we say so on the country page.

Verification

A record marked registry-verified appears in the named official register. Where multiple sources cover the same facility, fields are cross-checked and disagreements surface rather than being averaged away.

Update cadence

Countries re-ingest as their sources refresh. Each record carries its own vintage; the status page shows the newest vintage per country.

Attribution & upstream licences

Where a source carries a licence with attribution obligations, we state it here and preserve it per record, and licensed extracts carry it downstream:

Field dictionary

The 38 attributes in the canonical schema, grouped:

GroupFields
IdentityVeligraph ID, name, source record ID
Placecountry, region, district, ward, address, coordinates, geocoding confidence
Classificationfacility type, care level, ownership, ownership subtype
Capacityinpatient beds, inpatient and outpatient service, operating hours
Staffing13 cadres: doctors, nurses, midwives, pharmacists, lab scientists and technicians, dentists, CHOs, CHEWs, HIM officers, environmental health officers, attendants
Serviceson-site laboratory, imaging, pharmacy, ambulance, mortuary
Regulatorylicence status, registration status, operational status
Provenancesource name, vintage, licence, verification flags

FHIR mapping

The register maps onto HL7 FHIR R4: each facility is a Location with a managing Organization; on-site services map to HealthcareService resources. Veligraph IDs are carried as stable identifiers on all three, so FHIR consumers can join back to the portal and API. See the API docs.

Corrections

Run a facility and see an error? Email hello@veligraph.com with the facility's Veligraph ID (on every record page). Corrections that change a published figure are noted on the record.

Limitations